Flexible decision making under uncertainty for intelligent mobility on-demand

Transcription

1 Flexible decision making under uncertainty for intelligent mobility on-demand Michael Rigby and Stephan Winter Department of Infrastructure Engineering The University of Melbourne, Australia Abstract. Interacting with dynamic ride sharing systems for ad-hoc travel is a complex spatio-temporal task. The uncertainty of service supply and demand in this constrained arena challenges the rigidity of traditional human-computer interfaces. Without knowledge of service potential or the effect of their own limiting constraints, clients may simply not find any matching ride. Previous work on an intuitive interface concept, launch pads, resolves this issue by providing clients with visual feedback during a novel 2-step negotiation. Whilst computationally valid, human understanding of the launch pad metaphor and its interaction design still has to be assessed to close the system s feedback loop. For this purpose usability testing of launch pads is proposed in a directed wayfinding scenario. Results of testing will allow tuning of the system towards validation of the proposed visualization. Keywords: Intelligent transportation systems, ad-hoc ride sharing, humancomputer interaction, spatial cognitive engineering 1 Introduction New sensing and communications technologies are inspiring the design of innovative, shared ride transportation systems [25, 12], e.g., autonomous vehicle ride sharing. Using novel location based services integrating advanced traveler information, users can search for and discover opportunities satisfying their intentions in space-time [19], moving us closer towards realization of an intelligent Mobility Internet [14]. Existing human-computer interfaces for ad-hoc ride sharing however are rigid. They rely on the full and a priori disclosure of trip information from persons to perform ride matching [6]. For a user (client), this presents two significant issues. Firstly, disclosure of trip information before knowing the system s capacity to respond presents a privacy issue. Secondly, without feedback a client may not be aware of the systems current state, e.g., during low service coverage, or the potential limitations of their request. Consequently they may make iterative, uninformed queries across space-time and sub-optimal choices. Due to the inherent variability of service supply and demand in ad-hoc ride sharing, travel planning in this complex and dynamic system can be uncertain.

2 2 Rigby and Winter To assist client reasoning and decision-making, we address both issues in previous work [20], developing a conceptual ride sharing user interface called OppRide. OppRide replaces the traditional requirement that all information be submitted in one step, with a novel 2-step negotiation. Using our approach, visual feedback describing a client s matching pick-up opportunities is communicated in the form of launch pads. The visualization is derived in response to a client s drop-off constraints only, allowing the subsequent pick-up choice to be flexible in both space and time (Fig. 1). This directed wayfinding approach [15] may also be extended to alternative interaction scenarios. Launch pads are quantitative representations of vehicle service potential drawn from time geography research. Aggregated from slabs [11] derived from each matching vehicle s network-time prism [13] according to a client s drop-off constraints (location and arrival time window), launch pads describe OppRide s global service offering which may come from multiple vehicles. Contrasting a client s successful negotiation using launch pads against a traditional approach using shortest path calculations, e.g. [5], the OppRide interaction design yields the same result given that in retrospect the launch pads derived will always intersect the requested pick-up. By revealing information incrementally a client s location privacy is preserved until a formal contract is made. The launch pad visualization facilitates a client s flexible decision making under uncertainty in a manner offering them choice a factor critical in mobility. The focus of this study is to determine to what degree the launch pad metaphor is understandable and usable by client users within ad-hoc ride sharing. Whilst the concept is valid computationally, we will seek to validate the OppRide interface design by conducting human usability testing. We hypothesize that a discrete point representation of launch pads is both easier to understand and use for pick-up choices in on-demand mobility compared to a continuous representation. Understanding the complexity of route choice behavior in mobility, the study presented in this paper is preliminary only and focuses on decision making based on the presentation of alternative visual interfaces. Fig. 1. OppRide feedback during the 2-step negotiation.

3 2 Feedback Flexible decision making for intelligent mobility on-demand 3 The OppRide interaction design is drawn from both transportation and spatial cognition research [20]. The negotiation constitutes a feedback loop [24], to provide clients with improved situation awareness via a user-centered design [4]. Compared to a traditional visualization (e.g., Fig. 2a), when launch pads are visualized as two-dimensional (point, polyline or polygon) features, a client s initial reasoning is binary: is their current, private location or accessible region within or outside the system s service offering? From here a decision can be made, e.g., wait at the current location or trade time for space and move to a more serviceable area (Fig. 2b). This simple approach already overcomes the case of a client failing to get any ride due to a lack of knowledge and matches. By considering a client s flexibility in space and time, the interface s time window approach provides a relevant choice set [22] whilst reducing display clutter [1]. Extending this further, enhancement of the launch pad visualization to a third dimension (point, polyline or polygon), may allow a client to adjust their intention within the constraints of the service offering. Drawing on route choice behavior research [3], additional factors, e.g., seating or fare, can be visualized using, e.g., a color ramp, to directly address a client s mobility needs (Fig. 2c,d)[21]. With added information a client can seek to maximize their potential individual utility [2], thus leading to optimal decisions. The effect of this added information and its cognitive loading [17] will be examined in our study. Towards minimizing OppRide s gulfs of evaluation and execution [16, 18], we now describe our usability test. Fig. 2. Example visualizations: traditional (a) 2D vehicle routes (point, polyline and discrete) and launch pads (b) 2D service area (polygon and discrete), (c) 3D seating (polygon and discrete), (d) 3D fare (polyline and continuous; cyan = low, magenta = high). Visualizations derived from two vehicles satisfying the client s drop-off. 3 Usability Testing According to seminal usability theory [7], testing of the OppRide interface design must consider the psychology of human-computer interaction together with the psychology of wayfinding. Whilst maps can be understood by humans [11],

4 4 Rigby and Winter understanding and use of the launch pad metaphor in navigation scenarios [15] needs to be gauged. For this purpose we will test 20 users from a range of profiles, deeming this sample size to be sufficient [23]. Integrating usability, user utility and cognitive loading, our research applies the PACMAD model [8], interpreting cognitive loading from a visualization and spatial cognition perspective. The usability test is performed using a prototype version of the OppRide interface, developed using the Google Maps API 1 for its interactive map functionality and capacity to overlay launch pads. The test will be conducted outdoors using a mobile device for reasons of relevance and interaction using a common touch screen interface. The test will survey participant preferences and choices using the point and polyline launch pad variants only (we exclude the polygonal representation in this preliminary study as the generalization is not directly comparable). Commencing with a short animation introducing the OppRide concept, the test then narrows participation with the following statement: You intend to travel using OppRide to a local shopping center 5 from now. You are one person and your luggage consists of a backpack only. You are standing at 15 Theodore Street, Flora Hill in Victoria. You have submitted your drop-off information using OppRide and have just received launch pads overlaid on a digital map describing your matching pick-up opportunities in the next 5-10 minutes. From this information, please browse, reason and choose a pick-up option by clicking your desired location on the map within the launch pads shown. The first task asks participants for their preference for either 2D (x,y) point or polyline launch pads in 10 different un-timed scenarios using the statement above. Scenarios visualize launch pads in different network geometries with varying levels of service coverage (refer [20]). In the next two tasks participants are asked to choose a pick-up location in 10 different scenarios and responses are timed. The first of these asks participants to choose a pick-up location from a 2D (x,y) point then a polyline launch pad. The second task uses 3D (x,y,color:fare) launch pads in the same scenarios, yet randomizes the ordering of the variants to minimize any bias arising from a previous choice. To assist decision making regarding access distance and time to launch pads in the 5 minute (lower bound) window, an additional spatial analysis (accessibility) tool is made available in the interface. This facility allows the test to focus on aspects of spatial cognition relating to visualization and choice rather than perceptions of space-time which are more complex, e.g., [10, 9]. The usability test s observations include: Preferences for point or polyline Decision times using 2D and 3D variants Use of zoom and pan map functions Observed behavior: interaction and confidence Participant assessment concluding the test Analysis will compare participant preferences for the launch pad variants against their decision times. 1

5 4 Closing the loop Flexible decision making for intelligent mobility on-demand 5 Results from the usability test are expected to commence closure of the OppRide feedback loop. By establishing preference for and understanding of the launch pad variants, it is expected that with the increased knowledge a client can make better decisions in ride sharing. Due to the dynamism of ride sharing, future work will consider the stability of launch pads in different network geometries and supply-demand conditions. From here a minimum bound (time horizon) can be ascertained for a client s interactions during the negotiation process (refer mental affordances [19]). Additional future work from this study includes the use and structuring of launch pad variants; catering for issues of scale and granularity using, e.g., map zoom levels, and finalizing and testing a full featured OppRide interface. Further behavioral issues relating to client flexibility in space-time, the effects of additional launch pad dimensions, e.g., seating, on individual utility and the use of OppRide in a mobile environment are deemed to be outside the scope of the current study, yet will be considered in future work. Results of the usability test and a technical poster describing the OppRide architecture and launch pads will be presented at the workshop. Acknowledgments. The authors would like to thank Antonio Krüger and Maria Vasardani for their contributions. Research in this paper has been supported by the Australian Research Council Linkage Project LP References 1. Andrienko, G., Andrienko, N., Bak, P., Keim, D., Wrobel, S.: Visual Analytics of Movement. Springer, Berlin Heidelberg (2013) 2. Ben-Akiva, M.E., Lerman, S.R.: Discrete Choice Analysis: Theory and Application to Travel Demand. The MIT Press, Cambridge, MA, USA (1985) 3. Bovy, P.H.L., Stern, E.: Route Choice: Wayfinding in Transport Networks. Kluwer Academic Publishers, Dordrecht (1990) 4. Endsley, M.R., Jones, D.G.: Designing for Situation Awareness: An Approach to User-Centered Design. CRC Press, Boca Raton, FL, USA, 2 edn. (2011) 5. Fabri, A., Recht, P.: On dynamic pickup and delivery vehicle routing with several time windows and waiting times. Transportation Research Part B: Methodological 40(4), (2006) 6. Furuhata, M., Dessouky, M., Ordez, F., Brunet, M.E., Wang, X., Koenig, S.: Ridesharing: The state-of-the-art and future directions. Transportation Research Part B: Methodological 57, (2013) 7. Green, T.R.G., Petre, M.: Usability analysis of visual programming environments: A cognitive dimensions framework. Journal of Visual Languages and Computing 7(2), (1996) 8. Harrison, R., Flood, D., Duce, D.: Usability of mobile applications: literature review and rationale for a new usability model. Journal of Interaction Science 1(1), 1 (2013)

6 6 Rigby and Winter 9. Howard, L., Javadi, A., Yu, Y., Mill, R., Morrison, L., Knight, R., Loftus, M., Staskute, L., Spiers, H.: The hippocampus and entorhinal cortex encode the path and euclidean distances to goals during navigation. Current Biology 24(12), (2014) 10. MacEachren, A.M.: Travel time as the basis of cognitve distance. Professional Geographer 32(1), (1980) 11. MacEachren, A.M.: How Maps Work: Representation, Visualization, and Design. The Guilford Press, New York, U.S.A. (1995) 12. Miller, H.J.: Beyond sharing: cultivating cooperative transportation systems through geographic information systems. Journal of Transport Geography 31, (2013) 13. Miller, H.J.: Activities in space and time. In: Hensher, D., Button, K., Haynes, K., Stopher, P. (eds.) Handbook of Transport Geography and Spatial Systems. pp Emerald Group Publishing (2004) 14. Mitchell, W.J., Borroni-Bird, C., Burns, L.D.: Reinventing the Automobile: Personal Urban Mobility for the 21st Century. The MIT Press, Cambride, MA, USA (2010) 15. Montello, D.R.: Navigation. In: Shah, P., Miyake, A. (eds.) The Cambridge Handbook of Visuospatial Thinking. pp Cambridge University Press, New York, NY, USA (2005) 16. Norman, D.A.: Cognitive engineering. In: Norman, D.A., Draper, S.W. (eds.) User Centered Systems Design. Lawrence Erlbaum Associates, Hillsdale, NJ, USA (1986) 17. Paas, F., Tuovinen, J.E., Tabbers, H., Van Gerven, P.W.M.: Cognitive load measurement as a means to advance cognitive load theory. Educational Psychologist 38(1), (2003) 18. Raubal, M.: Cognitive engineering for geographic information science. Geography Compass 3(3), (2009) 19. Raubal, M., Miller, H.J., Bridwell, S.A.: User-centred time geography for locationbased services. Geografiska Annaler: Series B, Human Geography 88B(4), (2004) 20. Rigby, M., Krüger, A., Winter, S.: An opportunistic client user interface to support centralized ride share planning. In: Proceedings of the 21st ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems. ACM (2013) 21. Rigby, M., Winter, S.: Enhancing launch pads for decision making in intelligent mobility on demand (Extended Abstract). In: Proceedings of the 17th International Conference on Intelligent Transportation Systems. IEEE (2014), Forthcoming 22. Tenbrink, T.: Relevance in spatial navigation and communication. In: Stachniss, C., Schill, K., Uttal, D. (eds.) Spatial Cognition VIII, Lecture Notes in Computer Science, vol. 7463, pp Springer Berlin Heidelberg (2012) 23. Turner, C.W., Lewis, J.R., Nielsen, J.: Determining usability test sample size. In: Karwowski, W. (ed.) International Encyclopedia of Ergonomics and Human Factors, vol. 3. CRC Press, Boca Raton, FL, USA, 2 edn. (1986) 24. Wiener, N.: Cybernetics: Or Control and Communication in the Animal and the Machine. MIT Press, Cambridge, MA, USA (1948) 25. Winter, S., Nittel, S.: Ad-hoc shared-ride trip planning by mobile geosensor networks. International Journal of Geographical Information Science 20(8), (2006)