A Balanced Assignment Mechanism for Online Taxi Recommendation

Size: px
Start display at page:

Download "A Balanced Assignment Mechanism for Online Taxi Recommendation"

Transcription

1 217 IEEE 18th International Conference on Mobile Data Management A Balanced Assignment Mechanism for Online Taxi Recommendation Guang Dai, Jianbin Huang, Stephen Manko Wambura, Heli Sun School of Computer Science and Technology, Xidian University, Xi an, China School of Software, Xidian University, Xi an, China Department of Computer Science and Technology, Xi an Jiaotong University, Xi an, China Abstract Majority of taxi recommender systems mainly focused on satisfaction of passengers without considering fairness in assignment of taxi drivers. In this paper we propose a balanced assignment mechanism for online taxi recommendation (BAMOTR). BAMOTR provides a mechanism for fair assignment of drivers at some locations with specific routes to pick up passengers and ensures a short waiting time for passengers. Fair assignment is intended to minimize the differences in income among the taxi drivers. Analysis shows out that fair assignment of drivers and shortening the time the passenger wait before pick up is a trade-off problem. In this paper, we set a regulatory factor that can adjust the trade-off between fair assignment of drivers and shortening of waiting time of passengers. We also propose an efficient range refinement algorithm to solve online taxi recommendation problem in BAMOTR. It is theoretically and experimentally proved that range refinement algorithm ensures the same recommendation result as brute-force algorithm, however it greatly reduces the time overhead. We validate the performances of BAMOTR with extensive evaluations. Experimental results show that BAMOTR achieve better recommendation fairness than compared approaches and guarantee a short waiting time for passengers to be picked up. I. INTRODUCTION Taxi recommender systems have emerged recently with the development of Global Positioning System (GPS), mobile communication technologies, sensor and wireless networking technologies. Taxi recommender systems are very important as they can contribute to several benefits such as suggesting right places for getting passengers, increasing revenue to taxi drivers, reducing waiting time, traffic jams as well as minimizing fuel consumption and hence increasing the number of trips the drivers can perform. Fundamentally, there are two suggested modes for taxi recommender systems. The first one is Grab Single Mode that a passenger request is sent to multiple taxis, and DiDi is a good example in this category. The second mode is Assignment Mode that a passenger request is assigned to a taxi, and Uber mode is a good example in this category. Tremendous evolution in technology has seen DiDi switching to Assignment Mode. This is attributed to the fact that in Grab Single Mode a request is sent to multiple taxis, as these taxis have likely equal chances to grab the passenger request, this might lead to unreasonable allocation of taxis. This situation might lead to several circumstances including the following. Firstly the assigned taxi might not be in an optimum position to pick up the passenger and hence increase waiting time for the passenger and driver time simultaneously. Secondly the other drivers who are in nearby positions might be deprived of the opportunity. The best solution to this problem is to apply Assignment Mode which assign a taxi to pick up a passenger according to some strategies, such as minimizing passenger waiting time. However, the question comes which taxi should be recommended to a passenger? Recommender systems focusing on public transportation service have been studied extensively [1] [6]. In reality, in order to reduce the waiting time of passengers, recommender systems usually recommend the taxi that is nearest to passengers. Liao and Lee [1], [2] aim to reduce the waiting time of passengers, but they have proposed systems which focus only on passengers, without taking into consideration the fairness in assignment of taxis, which may lead to big difference in income of drivers. The taxi-sharing systems proposed in [3], [4] are intended to maximize the profit of taxis by ridesharing, however, these systems can lead also to big difference in income of drivers, and a relatively increased waiting time for passengers. The works in [5], [6] do not also take into consideration fair assignment of drivers as they recommend some routes to taxi drivers, such that given routes might have potential passengers. In this paper we propose BAMOTR, a balanced assignment mechanism for online taxi recommendation, which can considerably bridge the gap in income between drivers and guarantee a short waiting time for passengers. We theoretically and experimentally prove that range refinement algorithm with the local shortest path algorithm (RRA-LSP) ensures the same recommendation result as brute-force algorithm, however it greatly reduces the time overhead. We evaluate the fairness in recommendation and shortening of waiting time by BAMOTR based on the real city network and historical taxi trajectories in Shanghai, China. Experimental results show that BAMOTR greatly reduce the differences in drivers income and guarantee a short waiting time for passengers when a proper regulatory factor is selected. The main contributions of the paper are as follows: We formulate an evaluation function which optimally and efficiently selects one taxi from a given group of competing taxis for each passenger request, and define an online taxi recommendation problem in BAMOTR. We propose two approaches to solve this problem. We analytically show that RRA-LSP enhances both great /17 $ IEEE DOI 1.119/MDM

2 reduction of the time overhead and provision of the same recommendation results as brute-force algorithm. We prove some theorems pertaining to RRA-LSP, these theorems guarantee that RRA-LSP have the same the recommendation results as brute-force algorithm. We design a balanced assignment mechanism to recommend taxis for a group of passengers while taking care of conflict resolution from multiple requests, which finds tradeoff between the fairness in assignment of drivers and the shortening of the waiting time of passengers. We provide evaluation results and the recommendation on fairness provided by BAMOTR and the efficiency produced by RRA-LSP based on extensive experiments. The rest of the paper is organized as follows. Section 2 discusses related work, highlighting the difference between BAMOTR and the existing ones. Section 3 formulates the online taxi recommendation problem in BAMOTR. Section 4 propose two approach to solve this problem. Section 5 presents a balanced assignment mechanism. Section 6 illustrates experimental results. We final conclude the paper in Section 7. II. RELATED WORK Recommender systems focused on public transportation service have been studied extensively [1] [12]. Generally, these recommender systems can be divided roughly into two categories. The first category recommend taxis according to dynamic requests. The second category recommend taxis according to historic trajectories from which mobility of patterns of passengers and pick-up behaviors of drivers can be mined. For the first category of recommender systems, authors in [1], [2], [7] only focus on passenger waiting time when recommending taxis. when a new request is received, Liao [1] recommends the nearest taxi to the pick up passenger. Lee et al. [2] considers factors such as traffic situation in addition to various traffic related factors, and try to minimize passenger waiting time. Seow et al. [7] focus on minimizing total passenger waiting time by concurrent dispatching multiple taxis and allowing taxis to exchange their booking assignments. The taxi-sharing systems proposed in [3], [4], [8], [13], which allow occupied taxis to pick up new passengers on the fly, promising that increase the income of these taxis by ridesharing. Several studies in the second category have similarly be conducted extensively. Sequence of pick-up points learned from the historic trajectories are recommended to drivers in [9] [11]. Systems proposed in [5], [6] provide taxi drivers with actual driving routes connecting these points. Qian et al. [12] recommend driving routes construction with connected road sections to drivers according to a route assignment mechanism. Our problem is closely related to the first category where we recommend taxis according to dynamic requests. Solutions proposed in [3], [4], [8] are not favorable and optimally applicable to our problem as they do not take into consideration on fair and balanced assignment of taxis as well differences in income of the drivers. Similarly existing taxi recommender systems [1], [2], [7] do not put consideration into factors we have outlined above. Therefore, we propose BAMOTR considering fairness in taxi recommendation. Qian et al. [12] propose fair assignment mechanism on recommending routes to drivers, hence making their work completely different from ours as we are focusing on fair assignment mechanism on requests by passengers. To the best of our knowledge, the problem of recommending taxis for a group of passengers considering fair assignment of taxis has seldom been studied in the existing taxi recommender systems. There are challenges that we have taken into consideration in finding tradeoff, for instance ensuring fair assignment of drivers while minimizing differences in their income and at the same time shortening waiting time of the passenger. If we recommend taxis by only considering the waiting time, then we can recommend the earliest arriving taxi to passengers, but the question is can we guarantee the fairness in assignment of drivers? On the other hand, if we recommend taxis by only considering the fairness in assignment of drivers, again comes a question can we recommend the taxi with the minimum travel time to passengers, and guarantee the shortening waiting time? In this paper, we propose a balanced assignment mechanism to address the above challenges. III. PROBLEM FORMULATION A. Basic Concepts Definition 1. A road network is a directed graph G =(I,R), where I is a set of intersections and R I I is a set of road sections. An intersection I i denotes a road junction or road end. An road section R k =(I i,i j ) R represents from its initial intersection I i to its final intersection I j. To ease the following discussions, we denote the initial and final intersections of road section R k as R k.s and R k.e, respectively. Definition 2. A path P =< R 1,R 2...R k > is a sequence of k road sections, where R i R and R i R j if i j. The consecutive road section must share an intersection, i.e., R i.e = R i+1.s where 1 i k. Definition 3. A request set Req = {req 1,req 2...req m } is a group of m passenger requests. A request req i is associated with four main properties, the departure location req i.s, the destination req i.d, the request time req i.t and the travel cost TC i generated by the request, which is calculated according to the shortest path from req i.s to req i.d. Taking Shanghai as an example, if the shortest path from req i.s to req i.d is 5 km, The starting taxi billing price within 3 kilometers is 14 yuan and after that the transportation expenses per kilometer is 2.5 yuan, total travel cost TC i =14+(5-3)*2.5=19 yuan. Definition 4. A taxi driver set D = {D 1,D 2...D n } is a group of n competing taxi drivers. A taxi driver D j is associated with two main properties, The current location D j.loc and the accumulated income AI j of the taxi driver. 13

3 Variable TABLE I. DESCRIPTION OF MAJOR NOTATIONS Description m, n the numbers of requests and taxi drivers respectively. req i,d j the i-th request and the j-th taxi driver respectively. a regulatory factor, [, 1]. t i,j the travel time from D j to req i, t i,j =SP i,j /v. D ear, the earliest taxi to arrive req i. D near the nearest taxi from req i. TC i the travel cost generated by req i AI j the accumulated income of D j. EV A i,j the EVA value of D j for req i SP i,j,d i,j, the shortest path, the Euclidean distance and the LSP i,j local shortest path from D j to req i respectively. the i-th narrowed subrange in RRA that a circle with G i,g final center at req i and a radius of r i, and the final narrowed subrange in RRA. i=1,2... AImin i the minimum accumulated income of dirvers in G i, i=,1,2... For example, if a taxi driver D j have picked up two passenger request req 1 and req 2 so far, the accumulated income AI j =TC 1 + TC 2. For the ease of reference, we summarize the commonly used notations throughout the paper in Table I. B. Evaluation Function With respect to the above discussion, we know that we need to find the tradeoff between the fair assignment of drivers and the shortening of waiting time of passenger by assigning a right driver to a passenger request. Therefore, we propose an evaluation function which takes into account two evaluation principles. 1. Trade-off Principle. The fair assignment of drivers which means that minimize the differences of drivers income, and the shortening of waiting time of passengers such that conditions remains favorable to both passengers and drivers. We should tradeoff between them according to the previous discussion. 2. Priority Principle. Business is customer oriented so we need to develop a strategy such that we can ensure the satisfaction of passengers. Otherwise, customers may be lost. Naturally, the priority principle means that the shortening of waiting time of the passengers have higher priority than the fair assignment of drivers. Therefore, for a given passenger request req i, we need to evaluate which driver D j (j (1...n) ) is best suited to be recommended to pick up the passenger. According to the above two principles, we define the EVA function as: EV A i,j =(1 )AI j + e Δti,j (1) where [, 1] and Δt i,j = t i,j t i,ear j (1...n). In Eq. 1, AI j represents the accumulated income of D j so far. t i,j represents the travel time from D j to req i.itis obvious that t i,ear = min(t i,j ). Thus, Δt i,j is an extra time caused by recommending D j rather than D ear. Apparently, Δt i,j, holds true when D j is the earliest arriving driver. Quite a few advanced travel time prediction techniques [14] (e.g., incorporating real time traffic conditions) can be applied to estimate the travel time. However, since the traffic prediction TABLE II. AN EXAMPLE OF ONLINE TAXI RECOMMENDATION PROBLEM IN BAMOTR AI j t,j Δt,j EV A,j D D * D D D is not a focus of this paper, we calculate travel time t i,j = SP i,j /v for the sake of simplicity. The exponential processing of Δt i,j realizes the priority principle where is a regulatory factor. When is chosen such that its value approaches to 1, it means that we more focus on waiting time. On the other hand, when its value approaches to, it means that we care more on the difference of drivers income. C. Problem Formulation Definition 5. Online taxi recommendation problem in BAMOTR is defined as following. For a given request req i, we aim to recommend the optimal taxi with the minimum EVA value to pick up the passenger. Therefore, For a request req i, our objective is: min : EV A i,j (2) D j,j [1,n] To simplify the discussion, the problem is illustrated via an example. We assume that there are five drivers and a request req and set =.7. The AI j and t,j of each driver is described in Table II. Thus we can calculate Δt,j = t,j t,ear where t,ear =3, and calculate the EVA value of each driver according to Eq. 1. We will recommend D 2 to pick up req according to Eq. 2. Although D 1 is the earliest arriving driver, it has a large accumulated income. For D 4, although it possesses the minimum accumulated income, it still has the maximum EVA value because it will let the passenger to wait extra 5 minutes. These fulfil the trade-off principle and priority principle. By analyzing the taxi trajectories collected in Shanghai, China, we have the following statistical results. The distribution and cumulative distribution function (CDF) of the time of passengers on a taxi are plotted in Fig. 1. From the figure, it can be seen that about 8% of passengers time on a taxi is less than 2 minutes. This means that most of passengers destinations are not far from their departure. The distribution and CDF of travel cost generated by requests are depicted in Fig. 2. We calculate the travel cost according to the Shanghai taxi billing. It is observed that almost 6% of passengers travel cost is less than 2 Yuan. The waiting time of passengers has been analyzed in [12], it shows that almost 4% of taxis pick up a passenger in less than 5 minutes. The standard deviation of about 25 driver s accumulated income per day are plotted in Fig. 3, from the figure, it can be clearly observed that as the number of requests increases, the difference in income of the drivers also increases. 14

4 (a) (b) Fig. 1. Distribution and CDF of the time of passengers on a taxi. (a) (b) Fig. 2. Distribution and CDF of the travel cost generated by passengers. A. Brute-force Algorithm IV. METHODOLOGY For a given request, our goal is to find a taxi with the minimum EVA value, and recommend it to pick up the passenger. For obtaining the taxi, intuitively, we should calculate the EVA value of all taxis. Thus we must find the shortest path from each taxi to the passenger according to Eq 1. Obviously, this is a single source shortest path problem. Therefore, brute-force algorithm initially find the shortest path from each taxi to the passenger by Dijkstra s algorithm [15], then calculate the EVA value of all taxis, and recommend the optimal taxi to pick up the passenger. With brute-force algorithm, we can get the optimal recommendation results, because we traverse all taxis to find the optimal taxi. However, brute-force algorithm have a great computational overhead, for each passenger request req i,in order to calculate the EVA value of taxis, it uses Dijkstra s algorithm on the entire road network. Intuitively, for req i, if D j is far away from it, D j is almost impossible to be recommended, because the EVA value of D j increases exponentially as Δt i,j. When t i,ear is fixed, a big t i,j will cause a big Δt i,j, and a small increase in Δt i,j will result in rapidly increasing in EVA value. Since these taxis that are far away from req i are almost impossible to be recommended, how do we avoid searching for these taxis? A simple idea is that we only consider taxis within a local range. B. Range Refinement Algorithm According to the above discussion, we can only consider taxis within a local range to avoid searching for taxis that are far away from req i. However, how do we determine the extent of the local range? Suppose we randomly select a local range, such as a local range that a circle with center at req i and a radius of 3 km, but the optimal recommendation results will not be guaranteed. Therefore, there remain one question in this case, whether there is a way to find a local range Fig. 3. Standard Deviation of driver s accumulated income. which guarantee the same recommendation results as bruteforce algorithm, the solution of which is provided by range refinement algorithm (RRA). RRA firstly narrows the range by iterating and gets a subrange, in which RRA recommend a taxi to the passenger. Specifically, The subrange narrowed by RRA include the optimal taxi, which will be recommended to pick up the passenger (it can be proved by Theorem 1 and Properties 1 and 2). Not only does RRA narrow the scope, but also ensures the same recommendation results as brute-force algorithm. Some theorems are introduced first to describe how to narrow the scope in RRA. Before introducing these theorems, we introduce some variables. For a given request req i,we assume G is the initial range (G is the original road network), AImin is the minimum accumulated income of all drivers within G, and AI ear is the accumulated income of D ear. We also assume in RRA. In fact, if =,it means that we only focus on the difference of drivers income. Thus, we just need to recommend a taxi with the minimum accumulated income each allocation, which can be found by a simple traversal. Theorem 1. In RRA, for a given request req i,wecangeta subrange G that a circle with center at req i and a radius of r. The taxi, only within the range G, may be the optimal taxi. r is defined as r = SP i,ear + v ln (AIear AI 1+ min ) (1 ) (3) Proof: We first calculate the EVA value of the earliest arriving taxi D ear EV A i,ear =(1 )AI ear +e (ti,ear ti,ear) =(1 )AI ear + we also calculate the EVA value of other taxi D j EV A i,j =(1 )AI j + e (ti,j ti,ear) (SP =(1 )AI j + e i,j SPi,ear)/v Generally, it is not easy to get SP i,j for each driver D j (j (1...n)) unless use Dijkstra s algorithm on the entire road network. We assume d i,j is a Euclidean Distance between D j and req i. Apparently, SP i,j d i,j. we can calculate the EVA value of D j with d i,j EV A i,j =(1 )AI j + e (di,j SPi,ear)/v Thus, EV A i,j EV A i,j. Equality holds if SP i,j = d i,j.if EV A i,j >EVA i,ear, wehaveev A i,j >EVA i,ear, and D j 15

5 is not the optimal taxi. We assume a taxi with AI min is D s, and assume d i,s is the Euclidean Distance between D s and req i. We can calculate the EVA value of D s with d i,s EV A i,s =(1 )AI min + e (di,s SPi,ear)/v we know that there is possible for EV A i,s EV A i,ear only when EV A i,s EV A i,ear, otherwise EV A i,s >EVA i,ear because EV A i,s EV A i,s and EV A i,s >EVA i,ear. When EV A i,s EV A i,ear, then (1 )AI min + e (di,s SPi,ear)/v (1 )AI ear + e (di,s SPi,ear)/v (1 )(AI ear AI min ) d i,s SP i,ear + v ln (1 )(AIear AI min ) Equality holds only when EV A i,s = EV A i,ear, and we assume (1 )(AIear AI min d = SP i,ear + v ln ) +1 If d i,s >d, then EV A i,s >EVA i,ear, and we can confirm that D s is definitely not the optimal taxi. Now we will prove that for any taxi D j, the following proposition is true. If d i,j > d, then EV A i,j > EVA i,ear for each D j (j (1...n)). For D j, we known AI j AI min, and if d i,j >d,wehave EV A i,j =(1 )AI j + e (di,j SPi,ear)/v (1 )AI min + e (di,j SPi,ear)/v > (1 )AI min + e (d SPi,ear)/v = EV A i,ear Due to EV A i,j EV A i,j, then EV A i,j >EVA i,ear Consequently, for any taxi D j (j (1...n)), if the Euclidean Distance d i,j between D j and req i is greater than d, D j is definitely not the optimal taxi. We set r = d = SP i,ear + v ln (AIear AI 1+ min ) (1 ) Therefore, the optimal taxi must be within this range G that a circle with center at req i and a radius of r. Through Theorem 1, we can get a subrange which include the optimal taxi. Move over, there are two properties in RRA. Property 1. The narrowed subrange G can be continuously narrowed iteration until AI i 1 min is equal to AIi min Proof: Through Theorem 1, we can get a subrange G from G. Generally, we can get G i+1 from G i, let AImin i is the minimum accumulated income of all drivers within G i, and assume the radius of G i is r i, then the radius of G i+1 is (1 )(AIear AI i min r i+1 = SP i,ear + v ln ) +1 Due to G i G i 1, therefore AImin i AIi 1 min. Then r i+1 r i = v (ln <= (1 )(AIear AI i min ) +1 (1 )(AIear AIi 1 min ln ) +1 ) Therefor r i+1 r i. Equality holds if and only if AImin i = AI i 1 min, otherwise, r i+1 < r i, it means that the subrange G i can be further narrowed to be G i+1. Therefore, when AImin i = AIi 1 min, G i+1 is the same as G i, and the subrange will converge. We will get the final narrowed subrange G final. Property 2. The shortest path SP i,opt from the optimal taxi D opt to req i must be within G final. Proof: In Theorem 1, we known that if d i,j > r,we have EV A i,j > EVA i,ear, and D j is definitely not the optimal taxi. In fact, we can prove that if SP i,j > r,we have EV A i,j >EVA i,ear. Similarly to the proof in Theorem 1, For each D j, we known AI j AImin, and if SP i,j >r, we have (SP EV A i,j =(1 )AI j + e i,j SPi,ear)/v (1 )AImin (SP + e i,j SPi,ear)/v > (1 )AImin + e (r SPi,ear)/v = EV A i,ear Thus EV A i,j > EVA i,ear, and D j is definitely not the optimal taxi. if combining with the Property 1, we known that if SP i,j > r final, we have EV A i,j > EVA i,ear. Therefore, if a taxi is the optimal taxi D opt, The shortest path SP i,opt r final, which means SP i,opt must be within G final. According to Theorem 1 and Properties 1 and 2, RRA can provide the shortest path for the optimal taxi to pick up the passenger. This means that RRA have the same recommendation results as brute-force algorithm. However, RRA has a premise that the value of SP i,ear and AI ear is known. To calculate these values, we have to find D ear firstly. Bruteforce algorithm can find D ear by using Dijkstra s algorithm on the entire road network. But it spends a lot of time for finding D ear. To avoid the time cost for calculating the value of SP i,ear and AI ear, we propose the following algorithm, which get a subrange that is larger than subrange narrowed by RRA, and also guarantees the same recommendation result as brute-force algorithm. Algorithm 1 LSP. Input: Request req i, Driver D i, Road Network G Output: LSP i,j 1: Get the x-axis x 1 and y-axis y 1 of req i and the x-axis x 2 and y-axis y 2 of D j ; 2: LSP i,j ; 3: x min min(x 1,x 2 ); 4: x max max(x 1,x 2 ); 5: y min min(y 1,y 2 ); 6: y max max(y 1,y 2 ); 7: while LSP i,j = do 8: x min x min C; //C is a constant. 9: x max x max + C; 1: y min y min C; 11: y max y max + C; 12: G l a local rectangular range surrounded by (x min,y min ), (x min,y max), (x max,y min ) and (x max,y max); 13: LSP i,j the local shortest path from D j to req i in G l ; 14: end while 15: return LSP i,j ; 16

6 let r r = LSP i,near + v ln 1+ (AI lmax AI (SP i,ear + v ln (AIear AI 1+ min ) (1 ) ) = min ) (1 ) Fig. 4. An example of LSP. 1) Range Refinement Algorithm with Local Shortest Path Algorithm (RRA-LSP): To simplify the discussion, we first introduce the local shortest path algorithm (LSP) via an example and Algorithm 1 outlines LSP. Fig. 4 shows that there is a request req i and a driver D j, our goal is to quickly find a connectivity path from D j to req i. In reality, urban road network generally has relatively strong connectivity. Thus, we can quickly find a connectivity path from D j to req i with a relatively large probability in a local road network, which is main idea of LSP. Therefore, we first get a local range G 1 (the red range) according to Algorithm 1 such that if there is connectivity path from D j to req i, then stop. But actually we do not get a connectivity path in G 1. Thus, we re-calculate a local range G 2 (the blue range) according to Algorithm 1. Fortunately, we can get a connectivity path from req i to D j in G 2, which is the local shortest path LSP i,j from D j to req i. RRA-LSP initially finds a local shortest path LSP i,near from D near to req i by using LSP, and get a local range that a circle with center at req i and a radius of LSP i,near. In fact, we can directly find the shortest path SP i,near from D near to req i by using some shortest path algorithm, such as Dijkstra s algorithm [15] and A* algorithm [16], but LSP is more efficiency than these algorithm in this problem because LSP can quickly find a local shortest path. We assume AI lmax is the maximum accumulated income of drivers in this local range. Then we have the following Theorem 2. Theorem 2. In RRA-LSP, for a given request req i, we can get a subrange G that a circle with center at req i and a radius of r. The taxi, only within the range G, may be the optimal taxi. r is defined as r = LSP i,near + v ln 1+ (AI lmax AI min ) (1 ) (4) Proof: Due to LSP i,near is a local shortest path from D near to req i and D near is not certainly the earliest arriving taxi D ear,wehavelsp i,near SP i,near and SP i,near SP i,ear respectively. Therefore, LSP i,near SP i,ear. Move over, AI lmax is the maximum accumulated income of drivers in a range that a circle with center at req i and a radius of LSP i,near. The range include D ear. Otherwise, SP i,ear > LSP i,near. It means D ear is not the earliest arriving taxi. Therefore, D ear is within the range, and we have AI lmax AI ear. In Eq 2, we replace SP i,ear and AI ear by LSP i,near and AI lmax, respectively. then r = LSP i,near + v ln 1+ (AI lmax AI min ) (1 ) Thus, r r. It means we expand the narrowed subrange. Therefore, the optimal taxi must be within this range G that a circle with center at req i and a radius of r. From Theorem 2, we know that RRA-LSP expand the final narrowed range, which may bring some extra time cost when using Dijkstra algorithm in the final narrowed range. But it avoids the time cost for finding D ear. The experimental results show that the time cost of RRA-LSP is almost 24% of that of brute-force algorithm, and get the same recommendation result as brute-force algorithm. In addition, it is seem obvious that Properties 1 and 2 is still established in RRA-LSP. Algorithm 2 outlines RRA-LSP. Algorithm 2 RRA-LSP. Input: A Request req i, Driver Set D={D i }, Road Network G Output: Result=(D opt,req i ) 1: Result null; 2: D near the taxi nearest to req i ; 3: AImin the minimum accumulated income of drivers in G; 4: LSP i,near calculate the local shortest path from D near to req i by LSP; 5: G local a circle with center at req i and a radius of LSP i,near ; 6: AI lmax the maximum accumulated income of drivers in G local ; 7: k 1; 8: AImin k -1; 9: while AImin k AIk 1 min do 1: if AImin k then 11: k k+1; 12: end if 13: r k calculate radius according to Theorem 2 and Property 1; 14: G k a circle with center at req i and a radius of r k ; 15: AImin k the minimum accumulated income of drivers in G k; 16: end while 17: G final G k ; 18: D sub D j within G final ;//D j D; 19: calculate the path P i,j from D j to req i by Dijkstra s algorithm on G final ;//D j D sub ; 2: D sub filt(d sub); // removing disconnected drivers from D sub ; 21: calculate the EV A i,j of D j according to Eq. 1; // D j D sub ; 22: D opt find the optimal taxi from D sub ; 23: Result (D opt,req i ); 24: AI opt AI opt + TC i ; // update the accumulated income of D opt. 25: D D \ D opt; 26: return Result; 2) Example: To better understand Algorithm 2, we use an example shown in Fig. 5 to illustrate the process of RRA-LSP. In the example, we assume =.7 and v = 5m/min. As shown in Fig. 5(a), there is a request and lots of taxis in G,we first calculate the local shortest path LSP near from the nearest taxi to the request by Algorithm 1. Then we can find AI lmax and AImin. We assume that the values of LSP near, AI lmax and AImin are 12, 32 and 6, respectively. Thus, we can calculate r 1 = 3561 according to Eq. 4, and get a subrange G 1 that a circle with center at the request and a radius of r 1 as shown in Fig. 5(b). We then find AImin 1, and assume AImin 1 = 2. Now we should find whether AI1 min and AI min 17

7 Fig. 6. An example of handling Conflict. are equal according to Property 1. if AImin 1 = AI min,we can not continue to narrow the scope, and G 1 is the final narrowed subrange, otherwise there is AImin 1 >AI min,we can get a smaller subrange. In this example, it seems obviously that AImin 1 = 2 >AI min =6, thus, we can continue to narrow G 1, and calculate r 2 = 318. Thus, we get a smaller subrange G 2 shown in Fig. 5(c). We then find AImin 2, and assume AImin 2 = 2. In this iteration, AI2 min = AI1 min = 2. Thus, we can not continue to narrow the scope, and G 2 is the final narrowed subrange G final. We final recommend a taxi with a specific route to pick up the request in G final. The taxi must be the optimal taxi D opt, and the specific route must be the shortest path SP opt, which is proved in Theorems 1 and 2 and Properties 1 and 2. However, for brute-force algorithm, it will search the entire road network G to find D opt. Therefore, Not only does RRA-LSP greatly reduces the time overhead, but also ensures the same recommendation results as bruteforce algorithm. V. TAXI ASSIGNMENT MECHANISM For a given request req i, we aim to recommend the optimal taxi to pick up the passenger. However, the reality is always that there are multiple requests which need to be assigned at the same time. When we recommend taxis to multiple requests at the same time, there may be a conflict that a taxi is recommended to multiple requests, provided that the taxi have the minimum EVA for these requests. As shown in Fig. 6(a), D 1 is recommended req 1 and req 2 at the same time. Taking into consideration the stated conflict, we propose a realistic rule to solve it. Realistic Rule for Handling Conflict. When a taxi is recommended to multiple requests, the rule find one with the maximum travel cost from these requests, and recommend the taxi to pick up this request. The reason is that if a taxi is Fig. 5. An example of RRA-LSP. recommended to multiple requests, then the taxi has the right to choose, therefore the driver will choose a request with the maximum travel cost because he or she can earn more money. Under this principle, as shown in Fig. 6(b), req 2 will be recommended to D 1. Generally speaking, each time the conflict occurs, we give priority in allocating the request with the maximum travel cost. This means that we can sort the multiple requests according to the travel cost in descending order, and then assign these requests in turn. Therefore, we design a taxi assignment mechanism, considering the conflict, which recommend a taxi for each request. The assignment problem has been addressed largely by mechanism designers [17], [18], and authors of [12] design a route assignment mechanism to assign routes to taxis. As discussed in [12], a fair assignment mechanism for drivers should meet some conditions. And here, our taxi assignment mechanism should meet the following conditions. Firstly we need to consider that there is a fairness in assignment of drivers to requests. Secondly we should try to minimum waiting time for passenger as already discussed earlier. Incorporating these conditions we design a balanced assignment mechanism to taxis through the following steps: 1) At the beginning of the assignment, the accumulated income of each driver D j D is set to AI j =. 2) At stage k, the requests are sorted on their travel cost in descending order. We then recommend taxis for these requests in turn by RRA-LSP. It should be noted that a taxi which have been assigned, will no longer participate in the next allocation. 3) After the assignment at stage k, the accumulated income of driver D j is updated as AI j = AI j + CT i when the request req i is picked up by the driver D j. If a driver not pick up any request, the accumulated income of the driver will not be updated. The mechanism can reduce the difference in drivers income and guarantee the shortening of waiting time of passengers, which is also verified by experimental results. VI. EXPERIMENTAL EVALUATION To verify that BAMOTR reduces the difference of drivers income, and guarantee the shortening of waiting time of passengers, we conduct comprehensive experiments. Each experiment randomly generated passenger requests and taxi 18

8 Fig. 7. SDI: =.1. Fig. 8. SDI: =.1. Fig. 9. SDI: =.1. Fig. 1. SDI: =.4. Fig. 11. SDI: =.7. Fig. 12. SDI: =.9. Fig. 13. SDI: =.99. Fig. 14. SDI: =.999. passenger satisfaction, less waiting time will lead to higher passenger satisfaction. B. Compared Methods Fig. 15. SDI: number of assignment: 1. TABLE III. EVALUATION CONFIGURATIONS The number of drivers 5 The number of requests 15 The regulatory factor.1,.1,.1,.4,.7,.9,.99,.999 The number of assignments 1, , 1 drivers in the intersection based on historical taxi trajectory dataset. We use 3km/h as the driving speed in the experiments and set the constant C in LSP as 5 miles because most of road section is shorter than 5m. A. Setting Road Network. The simulation is based on the road network of Shanghai, which contains about 2242 road intersections and road sections. Trajectories. The trajectories were collected in Shanghai, China from approximately 4 probe-taxis operating over a period of 126 days [19]. A trajectory consists of a sequence of points. Each point contains seven fields: ID, timestamp, longitude, latitude, speed, angle, and status. The meaning of the first six fields is well understood. The last field is the current status of a taxi, indicating vacant and 1 for occupied. Standard Deviation of Driver s Income (SDI). We compute (AIj AI) SDI by SDI = 2 n (j (1, 2...n)). SDI can be used to estimate how much difference in drivers income, which can reflect the recommendation fairness. Passenger Waiting Time. Waiting time is the duration of time from the moment a passenger makes a request to the time when the passenger gets a taxi. It reflects the degree of In this paper, The proposed method RRA-LSP is compared with three approaches. The first one is the brute-force algorithm, which have been discussed previously in the paper. The second approach is greedy algorithm, which recommend a taxi to each request in a local range considered as a circle with center at the request and a radius of 3 miles according to the balanced assignment mechanism proposed in BAMOTR. If there are not taxis within this local range or the path from each taxi within this local range to the request is disconnect, greedy algorithm will recommend the nearest (Euclidean distance) taxi to the request. It should be noted that when a taxi is recommended to a request, the path from the recommended taxi to the request should be provided to calculate passenger waiting time. The last approach is NTR (Nearest Taxi Recommendation method), which is proposed in literature [1]. NTR directly recommend the nearest vacant taxi to pick up a passenger, and is one of the classic representative of existing methods, such as [1], [2], [7], which focusing on reducing waiting time consequently resulting in higher passenger satisfaction, however it doesn t guarantee for the fairness in assignment of drivers. As we known, NTR is also widely used in reality, for example Didi and Uber. By comparing RRA-LSP with NTR, we can verify that BAMOTR achieve fairer recommendation results than these existing methods. There are other taxi recommendation methods that can not be compared with BAMOTR. Such as [3], [4], [8], as they allow occupied taxis to pick up new passengers on the fly as opposed to our solution. The method in [5], [9], [12] recommend a path to taxi by mining passenger mobility patterns from the history trajectories. They do not recommend a taxi according to a dynamic request, but recommend a path to a taxi for reducing taxi idle driving time when there is no request. Therefore, the focus of these methods is different from BAMOTR. 19

9 C. Evaluation Results 1) Evaluation on SDI: In the experiment, the parameter settings are shown in Table III. The results are depicted in Figs There are several observations that can be made from the figures. Firstly, Brute-force and RRA-LSP have the same SDI no matter how much is. This is attributable to the fact that RRA-LSP has the same recommendation results as Bruteforce, which is consistent with the proof of Theorems 1 and 2 and Properties 1 and 2. Secondly, the SDI of NTR increases rapidly with the number of assignments, and independent of the value of. More over, the SDI of RRA-LSP, Brutefroce and Greedy is smaller and smaller than that of NTR with the number of assignments. It is becuase that RRA-LSP, Brute-froce and Greedy recommend taxis according to the balanced assignment mechanism in BAMOTR, but NTR just focus on waiting time, without considering the fair assignment of drivers. Thirdly, when is closer to, the SDI of RRA- LSP is smaller and increases more slowly with the number of assignments, and when is closer to 1, the SDI of RRA-LSP is more closer to NTR. This is because we care much about the difference in drivers income when is closer to, and we focus much on waiting time when is closer to 1. Finally, the SDI of these methods is plotted in Fig. 15 when the number of assignments is 1, it is clearly observed that except NTR, the SDI of other methods increases with. Fig. 16. Average Waiting Time. Fig. 17. Distribution of waiting time of NTR. Fig. 18. Average Running Time. Fig. 19. Running Time: =.1. 2) Evaluation on Waiting Time: The experimental set up is the same and the number of assignments is set to 1. Thus, the waiting time is an average of 1 allocations. The results are plotted in Fig. 16. As in the previous experiment, Brute-force and RRA-LSP have the same result in waiting time. The reason is the same as what is described in the previous experiment. We observe that when is smaller, the average waiting time of RRA-LSP is longer than NTR, this is because we are concerned much on the difference of drivers income when is smaller. We also observe that the average waiting time of RRA-LSP and Greedy decreases with, and the average waiting time of RRA-LSP is smallest when is bigger than.6. This means that in order to obtain a shorter waiting time, we need to increasing, however from the last experiment, we know that SDI of RRA-LSP increases as increases. Thus, we should select a proper to find tradeoff between SDI and waiting time, from the expermental data it can be seen that the value of for this case is equal to.7. The distribution of the waiting time of NTR is plotted in Fig. 17. It is similar as the real distribution of waiting time encountered in [12]. 3) Evaluation on Running Time: In this experiment, we explore the average running time of 1 allocations of each method which is depicted in Fig. 18. We observe that RRA- LSP has the least running time when larger than.2. On average, RRA-LSP incurs 4%, 48% and 286% more efficiency than NTR, Greedy and Brute-force, respectively. It means that RRA-LSP greatly reduces the time overhead relative to Brute- Force. In more detail, we explore the running time varies with the number of assignment when =.7. The results are plotted in Fig. 19. We observe that as the number of assignment increases, the running time of brute-force algorithm increases quickly, unlike methods whose running times increase slowly. D. Scalability More insights of scalability are presented for four approaches in this section. SDI, waiting time and running time are evaluated by changing the number of taxi drivers and the number of passenger requests. 1) The number of taxi drivers: In this experiment, we fix the number of assignment is 1, the number of passenger requests is 15 and =.7, and change the number of taxi drivers. As shown in Fig. 2, SDI decreases with the number of taxi drivers for four approaches. For NTR, SDI decreases obviously with the number of taxi drivers. For other methods, there is a slight decrease. However, even when the number of taxi drivers is equal to 1, RRA-LSP and Brute-force still achieve 51% and 456% better recommendation fairness than Greedy and NTR, respectively. We observe that the average waiting time of these approaches decreases as the number of taxi drivers increases, and RRA-LSP and Brute-force has the least waiting time. We also observe that except NTR, the average running time of these methods decreases with the number of taxi drivers. In addition, when the number of taxi drivers is 1, RRA-LSP is 12%, 25% and 33% more efficiency than NTR, Greedy and Brute-force, respectively. 2) The number of passenger requests: The effect of the number of passenger requests is depicted in Fig. 21, as the number of passenger requests increases, the SDI of NTR increases quickly, but the SDI of other methods just have a little increases. When the number of passenger requests is 3, RRA-LSP and Brute-force achieve 31% and 617% better recommendation fairness than Greedy and NTR, respectively. We observe that the waiting time of all methods has a little increase with the number of passenger requests. We also observe that the average running time of these methods increases with the number of passenger requests. Moreover, when the number of passenger requests is 3, RRA-LSP is 11

10 (a) SDI (b) Average waiting time (c) Average running time Fig. 2. Evaluation on the number of drivers: =.7, number of assignment: 1, number of requests: 5. (a) SDI (b) Average waiting time (c) Average running time Fig. 21. Evaluation on the number of requests: =.7, number of assignment: 1, number of drivers: 5. -2%, 79% and 231% more efficiency than NTR, Greedy and Brute-force, respectively. VII. CONCLUSION In this paper, we propose BAMOTR, a balanced assignment mechanism for online taxi recommendation, which is motivated by majority of existing taxi recommender systems that mainly focused on satisfaction of passengers without considering fairness in assignment of taxi drivers. In BAMOTR, a balanced assignment mechanism is proposed to solve a balance problem on how to find tradeoff fair assignment of drivers and shortening of waiting time of passengers? In addition, BAMOTR propose RRA-LSP, an efficient method, which ensures the same recommendation result as brute-force algorithm, however it is much more efficiency than bruteforce algorithm. Analysis of experimental results shows that BAMOTR achieve better recommendation fairness for taxi drivers than compared methods, which is verified by SDI. In future work, we look forward to consider the real-time traffic conditions so that we can fine-tune our method to fit in dynamic real situations. REFERENCES [1] Z. Liao, Real-time taxi dispatching using global positioning systems, Communications of the ACM, vol. 46, no. 5, pp , 23. [2] D.-H. Lee, H. Wang, R. Cheu, and S. Teo, Taxi dispatch system based on current demands and real-time traffic conditions, Transportation Research Record: Journal of the Transportation Research Board, no. 1882, pp , 24. [3] S. Ma, Y. Zheng, and O. Wolfson, T-share: A large-scale dynamic taxi ridesharing service, in Data Engineering (ICDE), 213 IEEE 29th International Conference on. IEEE, 213, pp [4] C. Tian, Y. Huang, Z. Liu, F. Bastani, and R. Jin, Noah: a dynamic ridesharing system, in Proceedings of the 213 ACM SIGMOD International Conference on Management of Data. ACM, 213, pp [5] N. J. Yuan, Y. Zheng, L. Zhang, and X. Xie, T-finder: A recommender system for finding passengers and vacant taxis, IEEE Transactions on Knowledge and Data Engineering, vol. 25, no. 1, pp , 213. [6] M. Qu, H. Zhu, J. Liu, G. Liu, and H. Xiong, A cost-effective recommender system for taxi drivers, in Proceedings of the 2th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 214, pp [7] K. T. Seow, N. H. Dang, and D.-H. Lee, A collaborative multiagent taxi-dispatch system, IEEE Transactions on Automation Science and Engineering, vol. 7, no. 3, pp , 21. [8] B. Cao, L. Alarabi, M. F. Mokbel, and A. Basalamah, Sharek: a scalable dynamic ride sharing system, in th IEEE International Conference on Mobile Data Management, vol. 1. IEEE, 215, pp [9] J. Huang, X. Huangfu, H. Sun, H. Li, P. Zhao, H. Cheng, and Q. Song, Backward path growth for efficient mobile sequential recommendation, IEEE Transactions on Knowledge and Data Engineering, vol. 27, no. 1, pp. 46 6, 215. [1] Y. Ge, H. Xiong, A. Tuzhilin, K. Xiao, M. Gruteser, and M. Pazzani, An energy-efficient mobile recommender system, in Proceedings of the 16th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 21, pp [11] M. Zhang, J. Liu, Y. Liu, Z. Hu, and L. Yi, Recommending pick-up points for taxi-drivers based on spatio-temporal clustering, in Cloud and Green Computing (CGC), 212 Second International Conference on. IEEE, 212, pp [12] S. Qian, J. Cao, F. L. Mouël, I. Sahel, and M. Li, Scram: A sharing considered route assignment mechanism for fair taxi route recommendations, in Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 215, pp [13] Y. Huang, F. Bastani, R. Jin, and X. S. Wang, Large scale real-time ridesharing with service guarantee on road networks, Proceedings of the VLDB Endowment, vol. 7, no. 14, pp , 214. [14] J. Yuan, Y. Zheng, X. Xie, and G. Sun, Driving with knowledge from the physical world, in Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining. ACM, 211, pp [15] E. W. Dijkstra, A note on two problems in connexion with graphs, Numerische mathematik, vol. 1, no. 1, pp , [16] W. Zeng and R. Church, Finding shortest paths on real road networks: the case for a*, International journal of geographical information science, vol. 23, no. 4, pp , 29. [17] N. Nisan and A. Ronen, Algorithmic mechanism design, in Proceedings of the thirty-first annual ACM symposium on Theory of computing. ACM, 1999, pp [18] A. Abdulkadiroglu and T. Sönmez, School choice: A mechanism design approach, The American Economic Review, vol. 93, no. 3, pp , 23. [19] Suvnet-trace data, 111

Exploring the Patterns of Human Mobility Using Heterogeneous Traffic Trajectory Data

Exploring the Patterns of Human Mobility Using Heterogeneous Traffic Trajectory Data Exploring the Patterns of Human Mobility Using Heterogeneous Traffic Trajectory Data Jinzhong Wang April 13, 2016 The UBD Group Mobile and Social Computing Laboratory School of Software, Dalian University

More information

Where to Find My Next Passenger?

Where to Find My Next Passenger? Where to Find My Next Passenger? Jing Yuan 1 Yu Zheng 2 Liuhang Zhang 1 Guangzhong Sun 1 1 University of Science and Technology of China 2 Microsoft Research Asia September 19, 2011 Jing Yuan et al. (USTC,MSRA)

More information

VISUAL EXPLORATION OF SPATIAL-TEMPORAL TRAFFIC CONGESTION PATTERNS USING FLOATING CAR DATA. Candra Kartika 2015

VISUAL EXPLORATION OF SPATIAL-TEMPORAL TRAFFIC CONGESTION PATTERNS USING FLOATING CAR DATA. Candra Kartika 2015 VISUAL EXPLORATION OF SPATIAL-TEMPORAL TRAFFIC CONGESTION PATTERNS USING FLOATING CAR DATA Candra Kartika 2015 OVERVIEW Motivation Background and State of The Art Test data Visualization methods Result

More information

IMPROVING INFRASTRUCTURE FOR TRANSPORTATION SYSTEMS USING CLUSTERING

IMPROVING INFRASTRUCTURE FOR TRANSPORTATION SYSTEMS USING CLUSTERING IMPROVING INFRASTRUCTURE FOR TRANSPORTATION SYSTEMS USING CLUSTERING Presented By: Apeksha Aggarwal Research Scholar, C.S.E. Department, IIT Roorkee Supervisor: Dr. Durga Toshniwal Associate Professor,

More information

Time Series Data Cleaning

Time Series Data Cleaning Time Series Data Cleaning Shaoxu Song http://ise.thss.tsinghua.edu.cn/sxsong/ Dirty Time Series Data Unreliable Readings Sensor monitoring GPS trajectory J. Freire, A. Bessa, F. Chirigati, H. T. Vo, K.

More information

Exploring Urban Mobility from Taxi Trajectories: A Case Study of Nanjing, China

Exploring Urban Mobility from Taxi Trajectories: A Case Study of Nanjing, China Exploring Urban Mobility from Taxi Trajectories: A Case Study of Nanjing, China Yihong Yuan and Maël Le Noc Department of Geography, Texas State University, San Marcos, TX, U.S.A. {yuan, mael.lenoc}@txstate.edu

More information

Exploring Urban Mobility from Taxi Trajectories: A Case Study of Nanjing, China

Exploring Urban Mobility from Taxi Trajectories: A Case Study of Nanjing, China Exploring Urban Mobility from Taxi Trajectories: A Case Study of Nanjing, China Yihong Yuan, Maël Le Noc Department of Geography, Texas State University,San Marcos, TX, USA {yuan, mael.lenoc}@ txstate.edu

More information

Morning Time: 1 hour 30 minutes

Morning Time: 1 hour 30 minutes ADVANCED SUBSIDIARY GCE 477/0 MATHEMATICS (MEI) Decision Mathematics THURSDAY 2 JUNE 2008 Morning Time: hour 30 minutes *CUP/T44383* Additional materials: Printed Answer Book (enclosed) MEI Examination

More information

Surge Pricing and Labor Supply in the Ride- Sourcing Market

Surge Pricing and Labor Supply in the Ride- Sourcing Market Surge Pricing and Labor Supply in the Ride- Sourcing Market Yafeng Yin Professor Department of Civil and Environmental Engineering University of Michigan, Ann Arbor *Joint work with Liteng Zha (@Amazon)

More information

CIV3703 Transport Engineering. Module 2 Transport Modelling

CIV3703 Transport Engineering. Module 2 Transport Modelling CIV3703 Transport Engineering Module Transport Modelling Objectives Upon successful completion of this module you should be able to: carry out trip generation calculations using linear regression and category

More information

Area Classification of Surrounding Parking Facility Based on Land Use Functionality

Area Classification of Surrounding Parking Facility Based on Land Use Functionality Open Journal of Applied Sciences, 0,, 80-85 Published Online July 0 in SciRes. http://www.scirp.org/journal/ojapps http://dx.doi.org/0.4/ojapps.0.709 Area Classification of Surrounding Parking Facility

More information

Encapsulating Urban Traffic Rhythms into Road Networks

Encapsulating Urban Traffic Rhythms into Road Networks Encapsulating Urban Traffic Rhythms into Road Networks Junjie Wang +, Dong Wei +, Kun He, Hang Gong, Pu Wang * School of Traffic and Transportation Engineering, Central South University, Changsha, Hunan,

More information

A MODIFIED CELLULAR AUTOMATON MODEL FOR RING ROAD TRAFFIC WITH VELOCITY GUIDANCE

A MODIFIED CELLULAR AUTOMATON MODEL FOR RING ROAD TRAFFIC WITH VELOCITY GUIDANCE International Journal of Modern Physics C Vol. 20, No. 5 (2009) 711 719 c World Scientific Publishing Company A MODIFIED CELLULAR AUTOMATON MODEL FOR RING ROAD TRAFFIC WITH VELOCITY GUIDANCE C. Q. MEI,,

More information

Trip Distribution Modeling Milos N. Mladenovic Assistant Professor Department of Built Environment

Trip Distribution Modeling Milos N. Mladenovic Assistant Professor Department of Built Environment Trip Distribution Modeling Milos N. Mladenovic Assistant Professor Department of Built Environment 25.04.2017 Course Outline Forecasting overview and data management Trip generation modeling Trip distribution

More information

Changes in the Spatial Distribution of Mobile Source Emissions due to the Interactions between Land-use and Regional Transportation Systems

Changes in the Spatial Distribution of Mobile Source Emissions due to the Interactions between Land-use and Regional Transportation Systems Changes in the Spatial Distribution of Mobile Source Emissions due to the Interactions between Land-use and Regional Transportation Systems A Framework for Analysis Urban Transportation Center University

More information

Understanding Travel Time to Airports in New York City Sierra Gentry Dominik Schunack

Understanding Travel Time to Airports in New York City Sierra Gentry Dominik Schunack Understanding Travel Time to Airports in New York City Sierra Gentry Dominik Schunack 1 Introduction Even with the rising competition of rideshare services, many in New York City still utilize taxis for

More information

Urban Computing Using Big Data to Solve Urban Challenges

Urban Computing Using Big Data to Solve Urban Challenges Urban Computing Using Big Data to Solve Urban Challenges Dr. Yu Zheng Lead Researcher, Microsoft Research Asia Chair Professor at Shanghai Jiaotong University http://research.microsoft.com/en-us/projects/urbancomputing/default.aspx

More information

Path and travel time inference from GPS probe vehicle data

Path and travel time inference from GPS probe vehicle data Path and travel time inference from GPS probe vehicle data Timothy Hunter Department of Electrical Engineering and Computer Science University of California, Berkeley tjhunter@eecs.berkeley.edu Pieter

More information

Optimizing Roadside Advertisement Dissemination in Vehicular CPS

Optimizing Roadside Advertisement Dissemination in Vehicular CPS Optimizing Roadside Advertisement Dissemination in Vehicular CPS Huanyang Zheng and Jie Wu Computer and Information Sciences Temple University 1. Introduction Roadside Advertisement Dissemination Passengers,

More information

Measuring Social Functions of City Regions from Large-scale Taxi Behaviors

Measuring Social Functions of City Regions from Large-scale Taxi Behaviors Work in Progress workshop at PerCom 2 Measuring Social Functions of City Regions from Large-scale Taxi Behaviors Guande Qi, Xiaolong Li, Shijian Li, Gang Pan and Zonghui Wang Department of Computer Science

More information

Traffic Modelling for Moving-Block Train Control System

Traffic Modelling for Moving-Block Train Control System Commun. Theor. Phys. (Beijing, China) 47 (2007) pp. 601 606 c International Academic Publishers Vol. 47, No. 4, April 15, 2007 Traffic Modelling for Moving-Block Train Control System TANG Tao and LI Ke-Ping

More information

Vehicle Routing and Scheduling. Martin Savelsbergh The Logistics Institute Georgia Institute of Technology

Vehicle Routing and Scheduling. Martin Savelsbergh The Logistics Institute Georgia Institute of Technology Vehicle Routing and Scheduling Martin Savelsbergh The Logistics Institute Georgia Institute of Technology Vehicle Routing and Scheduling Part II: Algorithmic Enhancements Handling Practical Complexities

More information

Shortening Picking Distance by using Rank-Order Clustering and Genetic Algorithm for Distribution Centers

Shortening Picking Distance by using Rank-Order Clustering and Genetic Algorithm for Distribution Centers Shortening Picking Distance by using Rank-Order Clustering and Genetic Algorithm for Distribution Centers Rong-Chang Chen, Yi-Ru Liao, Ting-Yao Lin, Chia-Hsin Chuang, Department of Distribution Management,

More information

A 2-Approximation Algorithm for Scheduling Parallel and Time-Sensitive Applications to Maximize Total Accrued Utility Value

A 2-Approximation Algorithm for Scheduling Parallel and Time-Sensitive Applications to Maximize Total Accrued Utility Value A -Approximation Algorithm for Scheduling Parallel and Time-Sensitive Applications to Maximize Total Accrued Utility Value Shuhui Li, Miao Song, Peng-Jun Wan, Shangping Ren Department of Engineering Mechanics,

More information

Mapping Accessibility Over Time

Mapping Accessibility Over Time Journal of Maps, 2006, 76-87 Mapping Accessibility Over Time AHMED EL-GENEIDY and DAVID LEVINSON University of Minnesota, 500 Pillsbury Drive S.E., Minneapolis, MN 55455, USA; geneidy@umn.edu (Received

More information

Outline Network structure and objectives Routing Routing protocol protocol System analysis Results Conclusion Slide 2

Outline Network structure and objectives Routing Routing protocol protocol System analysis Results Conclusion Slide 2 2007 Radio and Wireless Symposium 9 11 January 2007, Long Beach, CA. Lifetime-Aware Hierarchical Wireless Sensor Network Architecture with Mobile Overlays Maryam Soltan, Morteza Maleki, and Massoud Pedram

More information

An Online Ride-Sharing Path Planning Strategy for Public Vehicle Systems

An Online Ride-Sharing Path Planning Strategy for Public Vehicle Systems 1 An Online Ride-Sharing Path Planning Strategy for Public Vehicle Systems Ming Zhu, Xiao-Yang Liu, and Xiaodong Wang, Fellow, IEEE arxiv:1712.09356v1 [cs.ai] 27 Dec 2017 Abstract As efficient traffic-management

More information

Detecting Origin-Destination Mobility Flows From Geotagged Tweets in Greater Los Angeles Area

Detecting Origin-Destination Mobility Flows From Geotagged Tweets in Greater Los Angeles Area Detecting Origin-Destination Mobility Flows From Geotagged Tweets in Greater Los Angeles Area Song Gao 1, Jiue-An Yang 1,2, Bo Yan 1, Yingjie Hu 1, Krzysztof Janowicz 1, Grant McKenzie 1 1 STKO Lab, Department

More information

Human resource data location privacy protection method based on prefix characteristics

Human resource data location privacy protection method based on prefix characteristics Acta Technica 62 No. 1B/2017, 437 446 c 2017 Institute of Thermomechanics CAS, v.v.i. Human resource data location privacy protection method based on prefix characteristics Yulong Qi 1, 2, Enyi Zhou 1

More information

Greedy Algorithms. Kleinberg and Tardos, Chapter 4

Greedy Algorithms. Kleinberg and Tardos, Chapter 4 Greedy Algorithms Kleinberg and Tardos, Chapter 4 1 Selecting breakpoints Road trip from Fort Collins to Durango on a given route. Fuel capacity = C. Goal: makes as few refueling stops as possible. Greedy

More information

Revenue Maximization in a Cloud Federation

Revenue Maximization in a Cloud Federation Revenue Maximization in a Cloud Federation Makhlouf Hadji and Djamal Zeghlache September 14th, 2015 IRT SystemX/ Telecom SudParis Makhlouf Hadji Outline of the presentation 01 Introduction 02 03 04 05

More information

Leaving the Ivory Tower of a System Theory: From Geosimulation of Parking Search to Urban Parking Policy-Making

Leaving the Ivory Tower of a System Theory: From Geosimulation of Parking Search to Urban Parking Policy-Making Leaving the Ivory Tower of a System Theory: From Geosimulation of Parking Search to Urban Parking Policy-Making Itzhak Benenson 1, Nadav Levy 1, Karel Martens 2 1 Department of Geography and Human Environment,

More information

MULTIPLE CHOICE QUESTIONS DECISION SCIENCE

MULTIPLE CHOICE QUESTIONS DECISION SCIENCE MULTIPLE CHOICE QUESTIONS DECISION SCIENCE 1. Decision Science approach is a. Multi-disciplinary b. Scientific c. Intuitive 2. For analyzing a problem, decision-makers should study a. Its qualitative aspects

More information

Understanding taxi travel demand patterns through Floating Car Data Nuzzolo, A., Comi, A., Papa, E. and Polimeni, A.

Understanding taxi travel demand patterns through Floating Car Data Nuzzolo, A., Comi, A., Papa, E. and Polimeni, A. WestminsterResearch http://www.westminster.ac.uk/westminsterresearch Understanding taxi travel demand patterns through Floating Car Data Nuzzolo, A., Comi, A., Papa, E. and Polimeni, A. A paper presented

More information

Adaptive Contact Probing Mechanisms for Delay Tolerant Applications. Wang Wei, Vikram Srinivasan, Mehul Motani

Adaptive Contact Probing Mechanisms for Delay Tolerant Applications. Wang Wei, Vikram Srinivasan, Mehul Motani Adaptive Contact Probing Mechanisms for Delay Tolerant Applications Wang Wei, Vikram Srinivasan, Mehul Motani Outline Introduction Modeling contact processes Real world contact traces Designing energy

More information

Data-Driven Optimization under Distributional Uncertainty

Data-Driven Optimization under Distributional Uncertainty Data-Driven Optimization under Distributional Uncertainty Postdoctoral Researcher Electrical and Systems Engineering University of Pennsylvania A network of physical objects - Devices - Vehicles - Buildings

More information

Extracting mobility behavior from cell phone data DATA SIM Summer School 2013

Extracting mobility behavior from cell phone data DATA SIM Summer School 2013 Extracting mobility behavior from cell phone data DATA SIM Summer School 2013 PETER WIDHALM Mobility Department Dynamic Transportation Systems T +43(0) 50550-6655 F +43(0) 50550-6439 peter.widhalm@ait.ac.at

More information

Channel Allocation Using Pricing in Satellite Networks

Channel Allocation Using Pricing in Satellite Networks Channel Allocation Using Pricing in Satellite Networks Jun Sun and Eytan Modiano Laboratory for Information and Decision Systems Massachusetts Institute of Technology {junsun, modiano}@mitedu Abstract

More information

Modelling, Simulation & Computing Laboratory (msclab) Faculty of Engineering, Universiti Malaysia Sabah, Malaysia

Modelling, Simulation & Computing Laboratory (msclab) Faculty of Engineering, Universiti Malaysia Sabah, Malaysia 1.0 Introduction Intelligent Transportation Systems (ITS) Long term congestion solutions Advanced technologies Facilitate complex transportation systems Dynamic Modelling of transportation (on-road traffic):

More information

DO TAXI DRIVERS CHOOSE THE SHORTEST ROUTES?

DO TAXI DRIVERS CHOOSE THE SHORTEST ROUTES? DO TAXI DRIVERS CHOOSE THE SHORTEST ROUTES? A Large-Scale Analysis of Route Choice Behavior of Taxi Drivers Using Floating Car Data in Shanghai Junyan Li Superviors: Juliane Cron M.Sc. (Technische Universität

More information

Supplementary Technical Details and Results

Supplementary Technical Details and Results Supplementary Technical Details and Results April 6, 2016 1 Introduction This document provides additional details to augment the paper Efficient Calibration Techniques for Large-scale Traffic Simulators.

More information

A NOTE ON A SINGLE VEHICLE AND ONE DESTINATION ROUTING PROBLEM AND ITS GAME-THEORETIC MODELS

A NOTE ON A SINGLE VEHICLE AND ONE DESTINATION ROUTING PROBLEM AND ITS GAME-THEORETIC MODELS ALS Advanced Logistic Systems A NOTE ON A SINGLE VEHICLE AND ONE DESTINATION ROUTING PROBLEM AND ITS GAME-THEORETIC MODELS Andrzej Grzybowski Czestochowa University of Technology, Poland Abstract: In the

More information

Exploring Human Mobility with Multi-Source Data at Extremely Large Metropolitan Scales. ACM MobiCom 2014, Maui, HI

Exploring Human Mobility with Multi-Source Data at Extremely Large Metropolitan Scales. ACM MobiCom 2014, Maui, HI Exploring Human Mobility with Multi-Source Data at Extremely Large Metropolitan Scales Desheng Zhang & Tian He University of Minnesota, USA Jun Huang, Ye Li, Fan Zhang, Chengzhong Xu Shenzhen Institute

More information

We Are on the Way: Analysis of On-Demand Ride-Hailing Systems

We Are on the Way: Analysis of On-Demand Ride-Hailing Systems Vol., No., Xxxxx, pp. issn - eissn - 1 We Are on the Way: Analysis of On-Demand Ride-Hailing Systems Guiyun Feng Department of Industrial and Systems Engineering, University of Minnesota, Minneapolis,

More information

Social and Technological Network Analysis: Spatial Networks, Mobility and Applications

Social and Technological Network Analysis: Spatial Networks, Mobility and Applications Social and Technological Network Analysis: Spatial Networks, Mobility and Applications Anastasios Noulas Computer Laboratory, University of Cambridge February 2015 Today s Outline 1. Introduction to spatial

More information

CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability

CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability CS188: Artificial Intelligence, Fall 2009 Written 2: MDPs, RL, and Probability Due: Thursday 10/15 in 283 Soda Drop Box by 11:59pm (no slip days) Policy: Can be solved in groups (acknowledge collaborators)

More information

Nearest Neighbor Search with Keywords in Spatial Databases

Nearest Neighbor Search with Keywords in Spatial Databases 776 Nearest Neighbor Search with Keywords in Spatial Databases 1 Sphurti S. Sao, 2 Dr. Rahila Sheikh 1 M. Tech Student IV Sem, Dept of CSE, RCERT Chandrapur, MH, India 2 Head of Department, Dept of CSE,

More information

Texas A&M University

Texas A&M University Texas A&M University CVEN 658 Civil Engineering Applications of GIS Hotspot Analysis of Highway Accident Spatial Pattern Based on Network Spatial Weights Instructor: Dr. Francisco Olivera Author: Zachry

More information

Session-Based Queueing Systems

Session-Based Queueing Systems Session-Based Queueing Systems Modelling, Simulation, and Approximation Jeroen Horters Supervisor VU: Sandjai Bhulai Executive Summary Companies often offer services that require multiple steps on the

More information

Load Balancing in Distributed Service System: A Survey

Load Balancing in Distributed Service System: A Survey Load Balancing in Distributed Service System: A Survey Xingyu Zhou The Ohio State University zhou.2055@osu.edu November 21, 2016 Xingyu Zhou (OSU) Load Balancing November 21, 2016 1 / 29 Introduction and

More information

The design of Demand-Adaptive public transportation Systems: Meta-Schedules

The design of Demand-Adaptive public transportation Systems: Meta-Schedules The design of Demand-Adaptive public transportation Systems: Meta-Schedules Gabriel Teodor Crainic Fausto Errico ESG, UQAM and CIRRELT, Montreal Federico Malucelli DEI, Politecnico di Milano Maddalena

More information

Parking Slot Assignment Problem

Parking Slot Assignment Problem Department of Economics Boston College October 11, 2016 Motivation Research Question Literature Review What is the concern? Cruising for parking is drivers behavior that circle around an area for a parking

More information

Optimizing Roadside Unit (RSU) Placement in Vehicular CPS

Optimizing Roadside Unit (RSU) Placement in Vehicular CPS Optimizing Roadside Unit (RSU) Placement in Vehicular CPS Jie Wu Computer and Information Sciences Temple University RSU Placement l Roadside advertisement Attracting shoppers Variation of maximum coverage

More information

SIMC NUS High School of Mathematics and Science May 28, 2014

SIMC NUS High School of Mathematics and Science May 28, 2014 SIMC 014 NUS High School of Mathematics and Science May 8, 014 Introduction Singapore is world-renowned for her world-class infrastructure, including a highly developed transport system. Relative to her

More information

Massachusetts Institute of Technology

Massachusetts Institute of Technology .203J/6.28J/3.665J/5.073J/6.76J/ESD.26J Quiz Solutions (a)(i) Without loss of generality we can pin down X at any fixed point. X 2 is still uniformly distributed over the square. Assuming that the police

More information

Urban Population Migration Pattern Mining Based on Taxi Trajectories

Urban Population Migration Pattern Mining Based on Taxi Trajectories Urban Population Migration Pattern Mining Based on Taxi Trajectories ABSTRACT Bing Zhu Tsinghua University Beijing, China zhub.daisy@gmail.com Leonidas Guibas Stanford University Stanford, CA, U.S.A. guibas@cs.stanford.edu

More information

Lecture Note 1: Introduction to optimization. Xiaoqun Zhang Shanghai Jiao Tong University

Lecture Note 1: Introduction to optimization. Xiaoqun Zhang Shanghai Jiao Tong University Lecture Note 1: Introduction to optimization Xiaoqun Zhang Shanghai Jiao Tong University Last updated: September 23, 2017 1.1 Introduction 1. Optimization is an important tool in daily life, business and

More information

HEALTHCARE. 5 Components of Accurate Rolling Forecasts in Healthcare

HEALTHCARE. 5 Components of Accurate Rolling Forecasts in Healthcare HEALTHCARE 5 Components of Accurate Rolling Forecasts in Healthcare Introduction Rolling Forecasts Improve Accuracy and Optimize Decisions Accurate budgeting, planning, and forecasting are essential for

More information

Simulation of Process Scheduling Algorithms

Simulation of Process Scheduling Algorithms Simulation of Process Scheduling Algorithms Project Report Instructor: Dr. Raimund Ege Submitted by: Sonal Sood Pramod Barthwal Index 1. Introduction 2. Proposal 3. Background 3.1 What is a Process 4.

More information

Parking Occupancy Prediction and Pattern Analysis

Parking Occupancy Prediction and Pattern Analysis Parking Occupancy Prediction and Pattern Analysis Xiao Chen markcx@stanford.edu Abstract Finding a parking space in San Francisco City Area is really a headache issue. We try to find a reliable way to

More information

Research Article An Optimization Model of the Single-Leg Air Cargo Space Control Based on Markov Decision Process

Research Article An Optimization Model of the Single-Leg Air Cargo Space Control Based on Markov Decision Process Applied Mathematics Volume 2012, Article ID 235706, 7 pages doi:10.1155/2012/235706 Research Article An Optimization Model of the Single-Leg Air Cargo Space Control Based on Markov Decision Process Chun-rong

More information

LABELING RESIDENTIAL COMMUNITY CHARACTERISTICS FROM COLLECTIVE ACTIVITY PATTERNS USING TAXI TRIP DATA

LABELING RESIDENTIAL COMMUNITY CHARACTERISTICS FROM COLLECTIVE ACTIVITY PATTERNS USING TAXI TRIP DATA LABELING RESIDENTIAL COMMUNITY CHARACTERISTICS FROM COLLECTIVE ACTIVITY PATTERNS USING TAXI TRIP DATA Yang Zhou 1, 3, *, Zhixiang Fang 2 1 Wuhan Land Use and Urban Spatial Planning Research Center, 55Sanyang

More information

CHAPTER 3. CAPACITY OF SIGNALIZED INTERSECTIONS

CHAPTER 3. CAPACITY OF SIGNALIZED INTERSECTIONS CHAPTER 3. CAPACITY OF SIGNALIZED INTERSECTIONS 1. Overview In this chapter we explore the models on which the HCM capacity analysis method for signalized intersections are based. While the method has

More information

Delay management with capacity considerations.

Delay management with capacity considerations. Bachelor Thesis Econometrics Delay management with capacity considerations. Should a train wait for transferring passengers or not, and which train goes first? 348882 1 Content Chapter 1: Introduction...

More information

Microeconomic Theory -1- Introduction

Microeconomic Theory -1- Introduction Microeconomic Theory -- Introduction. Introduction. Profit maximizing firm with monopoly power 6 3. General results on maximizing with two variables 8 4. Model of a private ownership economy 5. Consumer

More information

Scheduling of Frame-based Embedded Systems with Rechargeable Batteries

Scheduling of Frame-based Embedded Systems with Rechargeable Batteries Scheduling of Frame-based Embedded Systems with Rechargeable Batteries André Allavena Computer Science Department Cornell University Ithaca, NY 14853 andre@cs.cornell.edu Daniel Mossé Department of Computer

More information

Implementing Visual Analytics Methods for Massive Collections of Movement Data

Implementing Visual Analytics Methods for Massive Collections of Movement Data Implementing Visual Analytics Methods for Massive Collections of Movement Data G. Andrienko, N. Andrienko Fraunhofer Institute Intelligent Analysis and Information Systems Schloss Birlinghoven, D-53754

More information

Math 273a: Optimization

Math 273a: Optimization Math 273a: Optimization Instructor: Wotao Yin Department of Mathematics, UCLA Fall 2015 online discussions on piazza.com What is mathematical optimization? Optimization models the goal of solving a problem

More information

Fundamental Characteristics of Urban Transportation Services

Fundamental Characteristics of Urban Transportation Services Fundamental Characteristics of Urban Transportation Services Anton J. Kleywegt School of Industrial and Systems Engineering Georgia Institute of Technology Smart Urban Transportation Forum Institute for

More information

Travel Pattern Recognition using Smart Card Data in Public Transit

Travel Pattern Recognition using Smart Card Data in Public Transit International Journal of Emerging Engineering Research and Technology Volume 4, Issue 7, July 2016, PP 6-13 ISSN 2349-4395 (Print) & ISSN 2349-4409 (Online) Travel Pattern Recognition using Smart Card

More information

Caesar s Taxi Prediction Services

Caesar s Taxi Prediction Services 1 Caesar s Taxi Prediction Services Predicting NYC Taxi Fares, Trip Distance, and Activity Paul Jolly, Boxiao Pan, Varun Nambiar Abstract In this paper, we propose three models each predicting either taxi

More information

Outline. 15. Descriptive Summary, Design, and Inference. Descriptive summaries. Data mining. The centroid

Outline. 15. Descriptive Summary, Design, and Inference. Descriptive summaries. Data mining. The centroid Outline 15. Descriptive Summary, Design, and Inference Geographic Information Systems and Science SECOND EDITION Paul A. Longley, Michael F. Goodchild, David J. Maguire, David W. Rhind 2005 John Wiley

More information

Decentralized Stabilization of Heterogeneous Linear Multi-Agent Systems

Decentralized Stabilization of Heterogeneous Linear Multi-Agent Systems 1 Decentralized Stabilization of Heterogeneous Linear Multi-Agent Systems Mauro Franceschelli, Andrea Gasparri, Alessandro Giua, and Giovanni Ulivi Abstract In this paper the formation stabilization problem

More information

Bicriterial Delay Management

Bicriterial Delay Management Universität Konstanz Bicriterial Delay Management Csaba Megyeri Konstanzer Schriften in Mathematik und Informatik Nr. 198, März 2004 ISSN 1430 3558 c Fachbereich Mathematik und Statistik c Fachbereich

More information

III.A. ESTIMATIONS USING THE DERIVATIVE Draft Version 10/13/05 Martin Flashman 2005 III.A.2 NEWTON'S METHOD

III.A. ESTIMATIONS USING THE DERIVATIVE Draft Version 10/13/05 Martin Flashman 2005 III.A.2 NEWTON'S METHOD III.A. ESTIMATIONS USING THE DERIVATIVE Draft Version 10/13/05 Martin Flashman 2005 III.A.2 NEWTON'S METHOD Motivation: An apocryphal story: On our last trip to Disneyland, California, it was about 11

More information

The Scope and Growth of Spatial Analysis in the Social Sciences

The Scope and Growth of Spatial Analysis in the Social Sciences context. 2 We applied these search terms to six online bibliographic indexes of social science Completed as part of the CSISS literature search initiative on November 18, 2003 The Scope and Growth of Spatial

More information

A Cellular Automaton Model for Heterogeneous and Incosistent Driver Behavior in Urban Traffic

A Cellular Automaton Model for Heterogeneous and Incosistent Driver Behavior in Urban Traffic Commun. Theor. Phys. 58 (202) 744 748 Vol. 58, No. 5, November 5, 202 A Cellular Automaton Model for Heterogeneous and Incosistent Driver Behavior in Urban Traffic LIU Ming-Zhe ( ), ZHAO Shi-Bo ( ô ),,

More information

A weighted mean velocity feedback strategy in intelligent two-route traffic systems

A weighted mean velocity feedback strategy in intelligent two-route traffic systems A weighted mean velocity feedback strategy in intelligent two-route traffic systems Xiang Zheng-Tao( 向郑涛 ) and Xiong Li( 熊励 ) School of Management, Shanghai University, Shanghai 200444, China (Received

More information

1.225 Transportation Flow Systems Quiz (December 17, 2001; Duration: 3 hours)

1.225 Transportation Flow Systems Quiz (December 17, 2001; Duration: 3 hours) 1.225 Transportation Flow Systems Quiz (December 17, 2001; Duration: 3 hours) Student Name: Alias: Instructions: 1. This exam is open-book 2. No cooperation is permitted 3. Please write down your name

More information

Analytical investigation on the minimum traffic delay at a two-phase. intersection considering the dynamical evolution process of queues

Analytical investigation on the minimum traffic delay at a two-phase. intersection considering the dynamical evolution process of queues Analytical investigation on the minimum traffic delay at a two-phase intersection considering the dynamical evolution process of queues Hong-Ze Zhang 1, Rui Jiang 1,2, Mao-Bin Hu 1, Bin Jia 2 1 School

More information

Improving the travel time prediction by using the real-time floating car data

Improving the travel time prediction by using the real-time floating car data Improving the travel time prediction by using the real-time floating car data Krzysztof Dembczyński Przemys law Gawe l Andrzej Jaszkiewicz Wojciech Kot lowski Adam Szarecki Institute of Computing Science,

More information

Urban Link Travel Time Estimation Using Large-scale Taxi Data with Partial Information

Urban Link Travel Time Estimation Using Large-scale Taxi Data with Partial Information Urban Link Travel Time Estimation Using Large-scale Taxi Data with Partial Information * Satish V. Ukkusuri * * Civil Engineering, Purdue University 24/04/2014 Outline Introduction Study Region Link Travel

More information

Travel Time Calculation With GIS in Rail Station Location Optimization

Travel Time Calculation With GIS in Rail Station Location Optimization Travel Time Calculation With GIS in Rail Station Location Optimization Topic Scope: Transit II: Bus and Rail Stop Information and Analysis Paper: # UC8 by Sutapa Samanta Doctoral Student Department of

More information

Lecture 19: Common property resources

Lecture 19: Common property resources Lecture 19: Common property resources Economics 336 Economics 336 (Toronto) Lecture 19: Common property resources 1 / 19 Introduction Common property resource: A resource for which no agent has full property

More information

GIS Based Transit Information System for Metropolitan Cities in India

GIS Based Transit Information System for Metropolitan Cities in India PAPER REFERENCE NO.: PN-250 GIS Based Transit Information System for Metropolitan Cities in India Pal, Sarvjeet. a and Singh, Varun. b a M. Tech. (GIS & Remote Sensing); GIS Cell; Motilal Nehru National

More information

Deep Algebra Projects: Algebra 1 / Algebra 2 Go with the Flow

Deep Algebra Projects: Algebra 1 / Algebra 2 Go with the Flow Deep Algebra Projects: Algebra 1 / Algebra 2 Go with the Flow Topics Solving systems of linear equations (numerically and algebraically) Dependent and independent systems of equations; free variables Mathematical

More information

Applications of Binary Search

Applications of Binary Search Applications of Binary Search The basic idea of a binary search can be used in many different places. In particular, any time you are searching for an answer in a search space that is somehow sorted, you

More information

A Gentle Introduction to Reinforcement Learning

A Gentle Introduction to Reinforcement Learning A Gentle Introduction to Reinforcement Learning Alexander Jung 2018 1 Introduction and Motivation Consider the cleaning robot Rumba which has to clean the office room B329. In order to keep things simple,

More information

The Model Research of Urban Land Planning and Traffic Integration. Lang Wang

The Model Research of Urban Land Planning and Traffic Integration. Lang Wang International Conference on Materials, Environmental and Biological Engineering (MEBE 2015) The Model Research of Urban Land Planning and Traffic Integration Lang Wang Zhejiang Gongshang University, Hangzhou

More information

Time in Distributed Systems: Clocks and Ordering of Events

Time in Distributed Systems: Clocks and Ordering of Events Time in Distributed Systems: Clocks and Ordering of Events Clocks in Distributed Systems Needed to Order two or more events happening at same or different nodes (Ex: Consistent ordering of updates at different

More information

Taxi services modeling for decision making support

Taxi services modeling for decision making support Young Researchers Seminar 2013 Young Researchers Seminar 2011 Lyon, France, June 5-7 2013 DTU, Denmark, June 8-10, 2011 Taxi services modeling for decision making support Session 4A : Transport Economics,

More information

Spatial Data Science. Soumya K Ghosh

Spatial Data Science. Soumya K Ghosh Workshop on Data Science and Machine Learning (DSML 17) ISI Kolkata, March 28-31, 2017 Spatial Data Science Soumya K Ghosh Professor Department of Computer Science and Engineering Indian Institute of Technology,

More information

Instructor (Brad Osgood)

Instructor (Brad Osgood) TheFourierTransformAndItsApplications-Lecture26 Instructor (Brad Osgood): Relax, but no, no, no, the TV is on. It's time to hit the road. Time to rock and roll. We're going to now turn to our last topic

More information

Traffic Flow Theory & Simulation

Traffic Flow Theory & Simulation Traffic Flow Theory & Simulation S.P. Hoogendoorn Lecture 4 Shockwave theory Shockwave theory I: Introduction Applications of the Fundamental Diagram February 14, 2010 1 Vermelding onderdeel organisatie

More information

Name: Date: Period: #: Chapter 1: Outline Notes What Does a Historian Do?

Name: Date: Period: #: Chapter 1: Outline Notes What Does a Historian Do? Name: Date: Period: #: Chapter 1: Outline Notes What Does a Historian Do? Lesson 1.1 What is History? I. Why Study History? A. History is the study of the of the past. History considers both the way things

More information

Queueing Theory and Simulation. Introduction

Queueing Theory and Simulation. Introduction Queueing Theory and Simulation Based on the slides of Dr. Dharma P. Agrawal, University of Cincinnati and Dr. Hiroyuki Ohsaki Graduate School of Information Science & Technology, Osaka University, Japan

More information

Scalable Scheduling with Burst Mapping in IEEE e (Mobile) WiMAX Networks

Scalable Scheduling with Burst Mapping in IEEE e (Mobile) WiMAX Networks Scalable Scheduling with Burst Mapping in IEEE 802.16e (Mobile) WiMAX Networks Mukakanya Abel Muwumba and Idris A. Rai Makerere University P.O. Box 7062 Kampala, Uganda abelmuk@gmail.com rai@cit.mak.ac.ug

More information

Season Finale: Which one is better?

Season Finale: Which one is better? CS4310.01 Introduction to Operating System Spring 2016 Dr. Zhizhang Shen Season Finale: Which one is better? 1 Background In this lab, we will study, and compare, two processor scheduling policies via

More information

Analysis of the Tourism Locations of Chinese Provinces and Autonomous Regions: An Analysis Based on Cities

Analysis of the Tourism Locations of Chinese Provinces and Autonomous Regions: An Analysis Based on Cities Chinese Journal of Urban and Environmental Studies Vol. 2, No. 1 (2014) 1450004 (9 pages) World Scientific Publishing Company DOI: 10.1142/S2345748114500043 Analysis of the Tourism Locations of Chinese

More information

Modelling exploration and preferential attachment properties in individual human trajectories

Modelling exploration and preferential attachment properties in individual human trajectories 1.204 Final Project 11 December 2012 J. Cressica Brazier Modelling exploration and preferential attachment properties in individual human trajectories using the methods presented in Song, Chaoming, Tal

More information