State Estimation with Nonlinear Inequality Constraints Based on Unscented Transformation
|
|
- Darleen Henderson
- 5 years ago
- Views:
Transcription
1 14th International Conference on Information Fusion Chicago, Illinois, USA, Jul 5-8, 2011 State Estimation with Nonlinear Inequalit Constraints Based on Unscented Transformation Jian Lan Center for Information Engineering Science Research School of Electronics and Information Engineering Xi an Jiaotong Universit, Xi an, , P. R. China X. Rong Li Department of Electrical Engineering Universit of New Orleans New Orleans, LA 70148, U.S.A. Abstract The state of some practical dnamic sstems satisfies constraints, which can be utilized to improve the performance of state estimation. State estimation with nonlinear inequalit constraints is a challenging problem. Projection methods are widel used to solve this problem. In this paper, a projection method is formulated as a special nonlinear function. Based on this formulation, unconstrained estimated states can be easil projected into the feasible constraint region through unscented transformation (UT), and both the constrained mean and covariance can be obtained. An approach is proposed to reduce the complexit of this projection process and a closed-form solution is derived for one-dimensional constraints. To solve the problem that the constrained covariance ma be ill-conditioned in UT-based approaches, a modification of the covariance is also proposed to account for necessar uncertainties. Two illustrative scenarios for ground moving target tracing are simulated to demonstrate the effectiveness and the efficienc of the proposed approaches. Kewords: State estimation, nonlinear inequalit constraints, unscented transformation, target tracing. I. INTRODUCTION In practice, the state of some dnamic sstems satisfies some constraints due to several reasons [1], e.g., basic laws of phsics, mathematical descriptions and practical limitations. Such constraints are generall described as a combination of linear or nonlinear equalities or inequalities. Properl utilizing the information provided b the constraints can improve the accurac of state estimation [2]. However, incorporating the constraints into the estimation process is challenging, especiall when the constraints are expressed as nonlinear inequalities. Man approaches have been proposed for state estimation with constraints. The include the pseudo-observation, projection, reparameterization methods and others (see, e.g., [1, 3 5]). The reparameterization methods (see, e.g., [2, 6]) attempt to reparameterize the sstem so that the equalit constraint is naturall satisfied. These methods are not eas to appl to the sstem with nonlinear inequalit constraints, and are also beond the scope of this paper. In this paper, onl the first two methods are discussed for real-time applications. In the pseudo-observation method (e.g., [7, 8]), the equalit constraints are treated as perfect observations (without noise) Research supported in part b ARO through grant W911NF and ONR-DEPSCoR through grant N and then a tpical filter, e.g., Kalman filter (KF), is applied to obtain the estimation results. Although this method has several merits, e.g., optimalit and clear intuitive interpretations, it ma suffer from numerical problems caused b a singular covariance matrix of measurement error and an increased computational burden caused b augmenting measurements [9]. Further, it is difficult to appl this approach to the sstem with nonlinear inequalit constraints. The projection method (e.g., [1, 9 11]) projects the estimation results onto the constraint surface or into the feasible region. This method can be applied to sstems with equalit or inequalit constraints. For state estimation with nonlinear equalit constraints, [1] identified two tpes of constraints for the estimated states: Tpe I (those that act on the entire distribution) and Tpe II (those that act on the mean of the distribution), and further pointed out that the pseudoobservation method enforces neither tpe of constraints and that the projection method (adopted in that paper) enforces the first tpe of constraint onl. A two-stage approach was then proposed in [1] to ensure that the second tpe of constraint is also satisfied in the projection method. For a sstem with linear inequalit constraints, [12] projected the unconstrained state estimate (conditional mean) onto some constrained set b solving a convex quadratic programming problem. This approach can also be applied to nonlinear inequalit constraints b linearizing the nonlinear function around the unconstrained conditional mean. The moving horizon based estimators (MHE) [3] [4] have been proposed to naturall ensure that the estimates satisf constraints b solving a constrained optimization problem. In a non-recursive form, this approach is computationall inefficient and ma be difficult to use for real-time applications [15]. On the other hand, constraints also affect the conditional error covariance of the estimated state. This fact was considered in [1] and the constrained covariance was calculated as in the extended KF (EKF) when the projection function is differentiable. For the case that the state is constrained with hard bounds, [13] also proposed an approach to calculate the constrained mean and covariance based on the unscented transformation (UT) [14] where the sigma points that violate the inequalit are projected normall onto the boundar of feasible region. Although the mathematical form of this projection ISIF 33
2 method for ever sigma point is not given in [13] (illustrated b several figures instead), the adopted projection method is actuall a special case of the one in [12]. Approaches with a similar idea can also be found in [15] and [16]. In these papers, it was shown that incorporating the constraint information into both the estimated mean and covariance can improve estimation performance. Because obtaining the distribution of the constrained state conditioned on measurements is ver difficult, using UT in the projection methods to calculate the constrained mean and covariance seems attractive, as adopted b the above approaches. However, this approach faces two problems: (a) Projecting ever sigma point into the feasible constraint region ma be computationall inefficient, especiall for high dimensional state vectors and for the case where the projected point is derived b solving an optimization problem. (b) Directl using the projected sigma points to calculate the covariance ma mae it ill-conditioned, especiall for nonlinear inequalit constraints. This phenomenon will be illustrated b a simple example later. This paper attempts to solve these two problems. First, a general formulation is given to clearl show the projection process based on optimization methods for nonlinear inequalit constraints. Based on this formulation, UT is adopted to calculate the constrained mean and covariance. An approach is also proposed to relieve the computational burden, and a closed-form solution for one-dimensional constraints is also derived. For the second problem, a modification of the constrained covariance is proposed, and the connections between this method and some other approaches are also discussed. This paper is organized as follows. Section II presents the above approaches as the main contents of this paper. Section III investigates an illustrative scenario for ground moving target tracing and presents simulation results and discussions. Section IV provides conclusions. II. STATE ESTIMATION WITH NONLINEAR INEQUALITY CONSTRAINTS BASED ON UT The following model with constraints is considered: x = F x 1 +w 1 z = H x +v (1) L c (x ) U where is the time index, x the state vector, z the nois measurement vector, w the process noise, v the measurement noise, F the transition matrix of the state x, and H the measurement matrix. w 1 and v are assumed to be mutuall uncorrelated zero-mean Gaussian white sequences with covariance matrixes Q 1 and R respectivel. c (x ) is the constraint function and U and L are the upper and the lower bounds of c (x ) respectivel. The purpose is to estimate x given measurement sequence Z {z 1,...,z } based on model (1). Here the projection method is considered. Assume that the unconstrained estimation results are given, i.e., the unconstrained state estimate ˆx and the error covariance matrix P are provided (b using the Kalman filter or Kalman-tpe filters). One tpical projection method can be given as (Tpe I constraint) ˆx P = argmin ( ˆx ) W ( ˆx ) (2) s.t. L c () U wherew is an smmetric positive definite weighting matrix. Note that the projection can also be applied to the one-step prediction ˆx 1. If c () is linear in and L = U, (2) reduces to the projection approach proposed in [9]. If c () is linear with L U, (2) is actuall the same as the method proposed in [12]. In [9] and [12], it was proved that among all the possible filters with the projection method (2) for a linear c (), the filter that uses W = P 1 has the smallest estimation error covariance. It was also pointed out that if c () is nonlinear, those approaches can also be applied b linearizing c () around ˆx as in the EKF. In fact, (2) projects onl the unconstrained mean ˆx rather than the whole distribution into the feasible region. [13] proposed an approach to calculate both the constrained mean ˆx P and error covariance PP based on UT. In this approach, the sigma points are generated from the unconstrained mean and covariance first, and the points outside the feasible region are projected onto its boundaries. The final results are calculated based on the projected sigma points. However, the mathematical expression of the projection process is not given. To analze this approach, we propose a general formulation of the projection process based on the optimization method (2). Assume that the unconstrained ˆx and P are the mean and the covariance matrix of random vector x respectivel: ˆx = E{x } P = E{(x ˆx )( ) } where ( ) means the item before. In fact, x represents an random vectors with the first two moments equal to ˆx and P respectivel, although these random vectors ma have different distributions. Different from (2), we consider projecting all the possible realizations of x into the feasible region rather than just its mean ˆx. This projection process can be described as (3) x P = argmin ( x ) W ( x ) (4) s.t. L c () U Then the constrained mean and error covariance can be calculated as ˆx P = E{xP } P P = E{(xP (5) ˆxP )( ) } The approach in (4) can actuall project the distribution of the unconstrained state into the feasible region. However, directl appling (4) to obtain ˆx P and PP is difficult, since the specific distribution of the unconstrained state is hardl nown 34
3 even if its first two moments are given. To solve this problem, an approach based on UT is presented next. A. Projection Approach Based on UT Since onl the first two moments ˆx and P of x are given, UT is considered to solve the problem above. For this purpose, the following nonlinear function is defined to embod the projection process (4): x P p(x ) = argmin( x ) W ( x ) (6) s.t. L c () U Function p(x ) will be referred to as the projection function in the sequel. B this definition, random vector x P is just a nonlinear function p(x ) of another random vector x with nown first two moments. This ma be the main reason wh the UT was applied for projection in [13] [16]. However, p(x ) is a special nonlinear function that ma be nondifferentiable at some points and most realizations of x P will be on the boundar of the feasible region, which maes directl appling UT here for the final results questionable. This will be analzed later. Simpl speaing, this UT-based projection process can be described as 1) Generate sigma points {s i } based on ˆx and P. Here n is the dimension of ˆx. 2) Calculate s i,p = p(s i ) for i = 1,..., as in (6). 3) Calculate ˆx P and PP based on {si,p } as in the unscented filter [14]. Actuall, the projection function of [13] (shown b several figures) can be viewed as a reduced form of (6) withc () = and W = I. The reason is that in [13], the sigma points outside the feasible region constrained b the upper and the lower bounds are directl assigned with the corresponding bound values (with other points unchanged), which is actuall the solution of (6) for each sigma point with c () = and W = I (see [17] for more details). Approaches with similar ideas can also be found in [16]. However, as stated in Section I, this projection approach based on UT faces two problems: (a) Projecting ever sigma point into the feasible region b solving (6) ma be computationall inefficient. (b) Directl using the projected sigma points to calculate the covariance for this special projection function ma be illconditioned. The following two subsections attempt to solve these two problems respectivel. B. Simplification of the Projection Approach In this subsection, an approach is proposed to reduce the computational complexit for the UT-based projection method. Application of this approach will be demonstrated in the simulation section. As described above, this approach needs to project each sigma point into the feasible region b using (6). B representing a sigma point (vector) as s, (6) can be rewritten as s P = argmin ( s ) W ( s ) (7) s.t. L c () U It is well nown that if the constraints are linear equalities, this optimization problem can be solved easil. That is, for the following problem, the solution is [2, 17] s P = argmin ( s ) W ( s ) (8) s.t. C = d s P = s +M (d C s ) (9) where M = W 1 C (C W 1 C ) 1. Based on the above results, we consider solving the following optimization problem with linear inequalit constraints: s P = argmin ( s ) W ( s ) (10) s.t. L C U This can be decomposed into two steps: 1) Solve (8) for s P (d ) as a function of d. 2) Then solve d P = argmin d (s P (d ) s ) W (s P (d ) s ) s.t. L d U (11) And the final result is given as s P (dp ). B solving (8) and substituting (9) into (11) ields d P = argmin d (d C s ) M W M (d C s ) s.t. L d U (12) Then the final result becomes s P (d P ) = s +M (d P C s ) (13) A proof of the optimalit of s P (dp ) as the solution of (10) is given next. Proof: Without loss of generalit, assume s 0 be the optimal solution of (10), then d 0 = C s 0 satisfies the constraints, i.e., L d 0 U. For an possible vector s that satisfies C s = d 0, onl s0 = s +M (d 0 C s ) has the minimum ( s (d ) s ) W ( s(d ) s ) as given b (9), which means that s 0 = s0 = s +M (d 0 C s ), since s 0 is optimal. That is to sa, the optimal solution of (10) with condition C = d,l d U (equivalent to the condition L C U ) must have the following linear form: s (d ) = s +M (d C s ),L d U (14) 35
4 or it is not optimal otherwise. Substitution of (14) into (10) ields (replace with s (d )): s P = argmin ( s ) W ( s ) s.t. L C U s P = s (d P ), d P = argmin d [(s +M (d C s ) s ) W (s +M (d C s ) s )] s.t. L C (s +M (d C s )) U s P = s (d P ), d P = argmin d (M (d C s )) W (M (d C s )) s.t. L d U s P = s (d P ), d P = argmin d (d C s ) M W M (d C s ) s.t. L d U (15) where means is equivalent to, M is given in (9), and C (s +M (d C s )) = d. The last in (15) indicates the optimalit of s P (dp ) in (13) as the solution of (10). This completes the proof. Note that solving (12) ma be much easier than solving (10) directl since in most cases C is a fat matrix with more columns than rows. If C is one-dimensional, a closed-form solution of (10) can be easil obtained b solving (12) as (13) and U, C s U d P = L, C s L (16) C s, elsewhere. If c () is nonlinear in, one common projection method is to linearize it around s [9,12]: c () c (s )+Dc (s )( s ) (17) and then the optimization problem (7) can be solved approximatel as in (12). In (17), Dc (s ) is the gradient of function c ( ) evaluated at s. Remar 1: The constraints in some practical sstems are one-dimensional. For example, (a) In ground moving target tracing, the road constraints can be expressed as one-dimensional inequalities. (b) In extended (cluster or group) object tracing, the distance in position between two objects of interest within the cluster satisfies some constraint, which can be formulated as one-dimensional inequalit. Thus, the above solution of the projection problem with a onedimensional inequalit can be useful in practice, which will also be demonstrated in the simulation section. C. State Estimation with Inequalit Constraints Based on UT In this subsection, we consider the second problem pointed out in Subsection A: directl using the projected sigma points to calculate the covariance ma be ill-conditioned. The main reason is that the projection function p(x ) defined in (6) is not a nice nonlinear function e.g., it is not differentiable at some specific points. This problem is analzed next and an approach for covariance modification in the UT-based projection is proposed to solve it. Tae the case with c () = and W = I in (6) for illustration. Under this assumption, U and L are the upper and the lower bounds for respectivel. Based on UT, the sigma points outside the feasible region defined b those bounds are directl assigned to the corresponding bound values b solving (6), as done in [13]. Consider two cases about this projection: (a) Some of the sigma points are within the feasible region. (b) None of the sigma points are within the feasible region. The projection approach based on UT for these two cases are shown in Figures 1 and 2 respectivel. In these figures the state vector x is assumed two-dimensional and the constraints are the upper and the lower bounds of each entr since c () =. Sigma points {s i }5 are generated from ˆx and P. In Figure 1 (Case 1), the projection process is similar to that of [13]. In this case, onl s 2,s3,s4 which violate the constraints are normall projected onto the boundar of the feasible region (since W = I) to become s 2,P,s 3,P,s 4,P respectivel. Then the final results are calculated as ˆx P = w i s i,p (18) P P = w i (s i,p ˆx P )( ) where n = 2, w i is the weight assigned to s i (i = 1,...,5) in UT, and s i,p = s i for i = 1,...,5 in Figure 1. B this projection, both mean and covariance of the estimated state are modified b the constraints. That is actuall wh this approach is better than those projecting onl ˆx into the feasible region [13]. s 3 s 2 s 3,P s 2,P s 1 (s1,p ) s 4,P s 5 (s5,p ) Figure 1. The Projection Approach Based on UT (Case 1) sigma point, projected sigma point. The rectangle represents the constraints. s 4 36
5 However, directl appling this projection approach to Case 2 is questionable as shown in Figure 2. In this case, all sigma points violate the constraints and are projected onto the boundar of the feasible region (left edge of the rectangle). Thus, P P calculated based on (18) is ill-conditioned as s 3 s 2 [ ] P P = P s 1 s 4 s 5 or [ ] P 22 s 5,P s 2,P s 1,P s 4,P s 3,P Figure 2. The Projection Approach Based on UT (Case 2) sigma point, projected sigma point. The rectangle represents the constraints. (19) Using this P P for further estimation ma mae the estimation process diverge. In fact, x can be distributed all over the feasible region even if the sigma points are all outside the region as in Figure 2, that is, the projected sigma points can not represent the constrained distribution, especiall when hard constraints exist. This ma be because the projection function p(x ) defined in (6) is actuall not differentiable due to the hard constraints. In view of this, we consider modifing the calculation of P P. Here for PP, the sigma points are modified to include more covariance information lost in the standard projection. The new sigma points are generated as (based on (13)) ŝ i,p = s i +λm (d i,p C s i ) (20) where 0 λ 1 is a parameter selected for including more covariance information and will be discussed later. Then ˆx P and P P can be calculated as ˆx P = w i s i,p (21) P P = w i (ŝ i,p ˆx P )( ) For common nonlinear functions (e.g., differentiable at an point), the final mean and covariance can be calculated as in (18) based on UT (treat p(x) as the functions). Compared with (18), the modified (21) is based on following considerations: (a) The constrained means ˆx P calculated b (18) and (21) respectivel are the same. The projected sigma points s i,p are in the feasible region. If the region is convex, as the convex combination (with weightsw i > 0 and i w i = 1) of s i,p, ˆxP also satisfies the constraints. In this case, no modification needs to be made to ˆx P without an further information about the distribution of the constrained state. (b) The constrained covariance P P in (21) is different from that in (18). The modification here is just to include more covariance information. For the standard projection, the onl modification to the original sigma point s i is the term M (d i,p C s i ), as shown in (13), which is also the main reason wh the constrained covariance matrix becomes ill-conditioned. Thus, this term is scaled down b λ in (20) to include the lost covariance information. (c) Given more information about the distribution of the unconstrained estimated state, more realizations of x can be obtained, and the corresponding projected points can be calculated b (6). Based on these projected points, more accurate constrained estimation results can be obtained. However, this ma not be feasible for some practical applications since it is difficult to obtain that distribution. In this case, using the above approach is optional to solve the problem of ill-conditioned covariance matrix due to the UT-based projection method. Different values of λ have different meanings, as stated next. Remar 2: 1) λ = 1 maes ŝ i,p = s i,p and thus P P reduces to the standard covariance matrix obtained b the standard projection approach as in (18). In this case, the projection function is treated as a common nonlinear function and UT is directl applied. 2) λ = 0 ields ŝ i,p = s i and ˆx P = w i s i,p P P = w i (ŝ i,p ˆx P )( ) = = w i (s i ˆxP )( ) w i (s i ˆx + ˆx ˆx P )( ) = P +(ˆx P ˆx )( ) (22) This actuall maes this method equivalent to the disploff method proposed in [1]. This method actuall indicates: B moving the estimate to a different value, the filter ceases to propagate the conditional mean, and the covariance must increase from P to P P [1]. 3) 0 < λ < 1 maes the covariance modification in a case between the above two. Generall, λ needs to be designed properl for different specific cases with different projection functions to obtain satisfactor performance. This is also discussed in the following simulation section. III. SIMULATION RESULTS To illustrate the effectiveness of the proposed approach, two scenarios for ground moving target tracing were designed as follows: 37
6 (a) Scenario 1 (S1). The target was on a linear road satisfing the following linear constraint: R L C x R U x [p x,vx,p,v ] (23) where is time index, p x,p and vx,v are the positions and velocities of the target in the x and directions, and C [ ] is the constraint matrix. The target performed the CV (nearl constant velocit) motion along the road and the trajectories were uniforml distributed within the road. (b) Scenario 2 (S2). The target was on a circular road satisfing the following nonlinear constraint: R L c (x ) R U x [p x,vx,p,v,ω ] (24) where ω is the turn rate, the nonlinear constraint function is c (x ) (p x px c) 2 +(p p c) 2, (p x c,p c) is the coordinates of the center of the circular road, and other notations are the same as in (23). The target performed the CT (nearl constant turn rate) motion along the road and the trajectories were uniforml distributed within the road. In (23) and (24), R U and R L are the upper and the lower bounds of the road width respectivel. The measurements are nois positions of the target at each sampling time. The above two scenarios are designed to evaluate the proposed algorithm under linear and nonlinear constraints respectivel. Specificall, the corresponding parameters and filters are designed as follows. For S1, the CV model was adopted in the algorithms, given b (1) with x [ p x v x p v], Φ = diag[f, F], [ ] [ ] [ ] T T H =, F, Q s 4 /3 T 3 /2 T 3 /2 T 2 w N[0, Q 1 ], Q 1 = q 1 diag[q s, Q s ], v N[0, R] (25) In the simulation, the following parameters were used: R = 20 2 m 2, q 1 = 0.09 (m/s 2 ) 2 (q 1 = 0 for the true trajectories), R L = 0 m, R U = 10 m. For this scenario, three filters were compared: 1) Kalman filter (KF) without considering the constraints. 2) KF with the unconstrained estimated mean projected (KF Proj), in which W = P 1. 3) KF with our approach (with modified covariance) for projection (KF Proj MD), in which W = P 1. λ = 0, 0.5, 1 were selected respectivel for further discussions. And the above filters were initialized optimall using the first two measurements. For S2, the CT model with 5 states was adopted in the algorithms, given b (1) with [18] x [ p x v x p v ], ω w N[0, Q 2 ], sinω 1 1 T ω cosω 1T ω cosω 1 T 0 sinω 1 T 0 1 cosω F = 0 1 T sinω ω T ω 1 0, 0 sinω 1 T 0 cosω 1 T [ ] H =, v N[0, R] where T is the sampling period and Q 2 is [19] [ ] Q 2 = T q 2 (26) (27) This formulation was chosen because it is simple and it has been tested that the result is rather insensitive to this simplification [19]. In the simulation, the following parameters were adopted: R = 20 2 m 2, Q 2 = diag([0,0,0,0,1])(rad/s) 2 (Q 2 = 0 for the true trajectories), R L = 300 m, R U = 305 m, [p x c,p c] = [0,302.5] m. Three filters based on this nonlinear model were compared: 1) UF without considering the constraints. 2) UF with the unconstrained estimated mean projected (UF Proj), in which W = P 1. 3) UF with our approach (with modified covariance) for projection (UF Proj MD), in which W = P 1. λ = 0, 0.5 were selected for further discussions. The filter with λ = 1 was found to diverge in this scenario. The above filters were initialized optimall based on the first two measurements using a predicted turn rate (e.g., ω 0 = 0 for simplicit). Since the constraints are one-dimensional, the closed form (16) was adopted for projection in KF Proj, KF Proj MD, UF Proj and UF Proj MD. The comparison results are the RMSE (root mean square errors) over 300 Monte Carlo runs for the two scenarios, shown in the following figures. position RMSE (m) KF KF KF Proj MD (λ = 0) KF Proj MD (λ = 0.5) KF Proj MD (λ = 1) time (s) Figure 3. RMS errors of position (m) in S1 There is little difference for KF proj MD with different λ values. 38
7 velocit RMSE (m/s) KF KF Proj KF MD (λ = 0) KF MD (λ = 0.5) KF MD (λ = 1) time (s) Figure 4. RMS errors of velocit (m/s) in S1 There is little difference for KF proj MD with different λ values. position RMSE (m) UF Proj MD (λ = 0) MD (λ = 0.5) time (s) Figure 5. RMS errors of position (m) in S2 Generall, the simulation results demonstrate that the proposed projection approach based on UT is effective, and that utilizing the constraint information based on the projection approaches is effective to improve the estimation performance. In S1, the KF Proj MD filters (λ = 0,0.5,1) outperform the other filters. The performance of KF Proj MD is insensitive to differentλ values (with linear inequalit constraints). In S2, however, the performance of UF Proj MD is rather sensitive to λ. The projection approach based on UT without covariance modification (λ = 1) was found to diverge in S2. The corresponding results are not shown. Simulation results also show that UF Proj MD with λ = 0.5 and λ = 0 has the best and worst performance respectivel (discussions on λ are given in Remar 2). The comparison results illustrate that modifing the constrained covariance b selecting λ is important for UT-based estimation with nonlinear inequalit constraints as in S2, but not for linear inequalit constraints as in S1. Also, the proposed UT-based projection approach with covariance modification appears effective in incorporating the constraint information for state estimation. Further research will be focused on the determination of λ for tpical practical applications. IV. CONCLUSIONS A projection method for nonlinear inequalit constraints has been formulated as a special nonlinear function. Based on this formulation, an approach has been proposed to relieve the computational burden, and an analtic solution for onedimensional constraints has been derived. The computational complexit is largel reduced b this approach. To solve the problem for which the covariance matrix ma be illconditioned while using UT for projection with hard constraints, a UT-based projection approach with a modified covariance matrix has been proposed, along with discussions on the connections between this approach and some other approaches. Two scenarios for ground moving target tracing with linear and nonlinear inequalit constraints have been simulated to evaluate the proposed approaches. Simulation results demonstrate that the proposed approach appears effective. velocit RMSE (m/s) Figure 6. UF Proj MD (λ = 0) MD (λ = 0.5) time (s) RMS errors of velocit (m/s) in S2 REFERENCES [1] S. J. Julier and J. J. Laviola, On Kalman filtering with nonlinear equalit constraints, IEEE Transactions on Signal Processing, vol. 55, no. 6, pp , [2] R. J. Hewett, M. T. Heath, M. D. Butala, F. Kamalabadi, A Robust Null Space Method for Linear Equalit Constrained State Estimation, IEEE Transactions on Signal Processing, vol. 58, no. 8, pp , [3] C. V. Rao, J. B. Rawlings, and J. H. Lee, Constrained linear state estimation a moving horizon approach, Automatica, vol. 37, no. 10, pp , [4] C. V. Rao, J. B. Rawlings, and D. Q. Mane, Constrained state estimation for nonlinear discrete-time sstems: stabilit and moving horizon approximations, IEEE Transactions on Automatic Control, vol. 48, no. 2, pp , [5] D. Simon and D. L. Simon, Constrained Kalman filtering via densit function truncation for turbofan engine health estimation, International Journal of Sstem Science, vol. 41, no. 2, pp , [6] L. Marle, Attitude error representations for Kalman filtering, Journal of Guidance, Control, and Dnamics, vol. 26, no. 2, pp ,
8 [7] M. Tah and J. L. Speer, Target tracing problems subject to inematic constraints, IEEE Transactions on Automatic Control, vol. 35, no. 3, pp , [8] A. T. Alouani and W. D. Blair, Use of a inematic constraint in tracing constant speed, maneuvering targets, IEEE Transactions on Automatic Control, vol. 38, no. 7, pp , [9] D. Simon and T. L. Chia, Kalman filtering with state equalit constraints, IEEE Transactions on Aerospace and Electronic Sstems, vol. 38, no. 1, pp , [10] T. Chia, P. Chow, and H. J. Chizec, Recursive parameter identification of constrained sstems: An application to electricall stimulated muscle, IEEE Transactions on Biomedical Engineering, vol. 38, no. 5, pp , [11] C. Yang and E. Blasch, Kalman filtering with nonlinear state contraints, in The 9th International Conference on Information Fusion 2006, Florence, Ital, Jul, 2006, pp [12] D. Simon, D. L. Simon, Aircraft turbofan engine health estimation using constrained Kalman filtering, ASME Journal of Engineering for Gas Turbine and Power, vol. 127, pp , [13] R. Kandepu, B. Foss, and L. Imsland, Appling the unscented Kalman filter for nonlinear state estimation, Journal of Process Control, Vol. 18, no. 7 8, pp , [14] S. J. Julier, J. K. Uhlmann, and H. Durrant-Whte, A new method for the nonlinear transformation of means and covariances in filters and estimators. IEEE Transactions on Automatic Control, vol. 45, no. 6, pp , [15] P. Vachhani, S. Narasimhan and R. Rengaswam, Robust and reliable estimation via Unscented Recursive Nonlinear Dnamic Data Reconciliation, Journal of Process Control, vol. 16, no. 10, pp , [16] S. Kolas, B. A. Fossa and T. S. Schei, Constrained nonlinear state estimation based on the UKF approach, Computers and Chemical Engineering, vol. 33, no. 8, pp , [17] D. Simon, Optimal state estimation. Hoboen, NJ: Wile-Interscience, [18] X. R. Li and V. P. Jilov, Surve of Maneuvering Target Tracing Part I: Dnamic Models, IEEE Transactions on Aerospace and Electronic Sstems, vol. 39, no. 4, pp , [19] F. Gustafsson and A. J. Isasson, Best Choice of State Variables for Tracing Coordinated Turns, in The 35th IEEE Conference on Decision and Control, Kobe, Japan, Dec., 1996, pp
Constrained State Estimation Using the Unscented Kalman Filter
16th Mediterranean Conference on Control and Automation Congress Centre, Ajaccio, France June 25-27, 28 Constrained State Estimation Using the Unscented Kalman Filter Rambabu Kandepu, Lars Imsland and
More informationConstrained Target Motion Modeling Part I: Straight Line Track
Constrained Target Motion Modeling Part I: Straight Line Trac Zhansheng Duan Center for Information Engineering Science Research Xi an Jiaotong University Xi an, Shaanxi 749, China Email: zduan@uno.edu
More informationUnscented Filtering for Interval-Constrained Nonlinear Systems
Proceedings of the 47th IEEE Conference on Decision and Control Cancun, Mexico, Dec. 9-11, 8 Unscented Filtering for Interval-Constrained Nonlinear Systems Bruno O. S. Teixeira,LeonardoA.B.Tôrres, Luis
More informationKalman Filtering with State Constraints
Kalman Filtering with State Constraints How an optimal filter can get even better Dan Simon January 18, 2008 The Kalman filter is the optimal minimum-variance state estimator for linear dynamic systems
More informationA comparison of estimation accuracy by the use of KF, EKF & UKF filters
Computational Methods and Eperimental Measurements XIII 779 A comparison of estimation accurac b the use of KF EKF & UKF filters S. Konatowski & A. T. Pieniężn Department of Electronics Militar Universit
More informationRecursive LMMSE Filtering for Target Tracking with Range and Direction Cosine Measurements
Recursive Filtering for Target Tracing with Range and Direction Cosine Measurements Zhansheng Duan Yu Liu X. Rong Li Department of Electrical Engineering University of New Orleans New Orleans, LA 748,
More informationESMF Based Multiple UAVs Active Cooperative Observation Method in Relative Velocity Coordinates
Joint 48th IEEE Conference on Decision and Control and 8th Chinese Control Conference Shanghai, P.R. China, December 6-8, 009 WeCIn5.4 ESMF Based Multiple UAVs Active Cooperative Observation Method in
More informationComputers and Chemical Engineering
Computers and Chemical Engineering 33 (2009) 13861401 Contents lists available at ScienceDirect Computers and Chemical Engineering journal homepage: www.elsevier.com/locate/compchemeng Constrained nonlinear
More informationState Estimation of Linear and Nonlinear Dynamic Systems
State Estimation of Linear and Nonlinear Dynamic Systems Part III: Nonlinear Systems: Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) James B. Rawlings and Fernando V. Lima Department of
More informationExtension of the Sparse Grid Quadrature Filter
Extension of the Sparse Grid Quadrature Filter Yang Cheng Mississippi State University Mississippi State, MS 39762 Email: cheng@ae.msstate.edu Yang Tian Harbin Institute of Technology Harbin, Heilongjiang
More information1 Introduction ISSN
Techset Composition Ltd, Salisbury Doc: {IEE}CTA/Articles/Pagination/CTA58454.3d www.ietdl.org Published in IET Control Theory and Applications Received on 15th January 2009 Revised on 18th May 2009 ISSN
More informationROBUST CONSTRAINED ESTIMATION VIA UNSCENTED TRANSFORMATION. Pramod Vachhani a, Shankar Narasimhan b and Raghunathan Rengaswamy a 1
ROUST CONSTRINED ESTIMTION VI UNSCENTED TRNSFORMTION Pramod Vachhani a, Shankar Narasimhan b and Raghunathan Rengaswamy a a Department of Chemical Engineering, Clarkson University, Potsdam, NY -3699, US.
More informationA New Nonlinear Filtering Method for Ballistic Target Tracking
th International Conference on Information Fusion Seattle, WA, USA, July 6-9, 9 A New Nonlinear Filtering Method for Ballistic arget racing Chunling Wu Institute of Electronic & Information Engineering
More informationVISION TRACKING PREDICTION
VISION RACKING PREDICION Eri Cuevas,2, Daniel Zaldivar,2, and Raul Rojas Institut für Informati, Freie Universität Berlin, ausstr 9, D-495 Berlin, German el 0049-30-83852485 2 División de Electrónica Computación,
More informationEKF, UKF. Pieter Abbeel UC Berkeley EECS. Many slides adapted from Thrun, Burgard and Fox, Probabilistic Robotics
EKF, UKF Pieter Abbeel UC Berkeley EECS Many slides adapted from Thrun, Burgard and Fox, Probabilistic Robotics Kalman Filter Kalman Filter = special case of a Bayes filter with dynamics model and sensory
More informationEKF, UKF. Pieter Abbeel UC Berkeley EECS. Many slides adapted from Thrun, Burgard and Fox, Probabilistic Robotics
EKF, UKF Pieter Abbeel UC Berkeley EECS Many slides adapted from Thrun, Burgard and Fox, Probabilistic Robotics Kalman Filter Kalman Filter = special case of a Bayes filter with dynamics model and sensory
More informationFisher Information Matrix-based Nonlinear System Conversion for State Estimation
Fisher Information Matrix-based Nonlinear System Conversion for State Estimation Ming Lei Christophe Baehr and Pierre Del Moral Abstract In practical target tracing a number of improved measurement conversion
More informationState Estimation using Moving Horizon Estimation and Particle Filtering
State Estimation using Moving Horizon Estimation and Particle Filtering James B. Rawlings Department of Chemical and Biological Engineering UW Math Probability Seminar Spring 2009 Rawlings MHE & PF 1 /
More informationA New Nonlinear State Estimator Using the Fusion of Multiple Extended Kalman Filters
18th International Conference on Information Fusion Washington, DC - July 6-9, 2015 A New Nonlinear State Estimator Using the Fusion of Multiple Extended Kalman Filters Zhansheng Duan, Xiaoyun Li Center
More informationAircraft Turbofan Engine Health Estimation Using Constrained Kalman Filtering. Funded by the NASA Aviation Safety Program June 16, 2003
Aircraft Turbofan Engine Health Estimation Using Constrained Kalman Filtering Dan Simon Electrical Engineering Dept. Cleveland State University Cleveland, Ohio Donald L. Simon US Army Research Laboratory
More informationLinear programming: Theory
Division of the Humanities and Social Sciences Ec 181 KC Border Convex Analsis and Economic Theor Winter 2018 Topic 28: Linear programming: Theor 28.1 The saddlepoint theorem for linear programming The
More informationIdentification of Nonlinear Dynamic Systems with Multiple Inputs and Single Output using discrete-time Volterra Type Equations
Identification of Nonlinear Dnamic Sstems with Multiple Inputs and Single Output using discrete-time Volterra Tpe Equations Thomas Treichl, Stefan Hofmann, Dierk Schröder Institute for Electrical Drive
More informationACTIVE FAILURE DETECTION: AUXILIARY SIGNAL DESIGN AND ON-LINE DETECTION
ACTIVE FAILURE DETECTION: AUXILIARY SIGNAL DESIGN AND ON-LINE DETECTION R. Nikoukhah,S.L.Campbell INRIA, Rocquencourt BP 5, 7853 Le Chesna Cedex, France. fax: +(33-39-63-57-86 e-mail: ramine.nikoukhah@inria.fr
More informationin a Rao-Blackwellised Unscented Kalman Filter
A Rao-Blacwellised Unscented Kalman Filter Mar Briers QinetiQ Ltd. Malvern Technology Centre Malvern, UK. m.briers@signal.qinetiq.com Simon R. Masell QinetiQ Ltd. Malvern Technology Centre Malvern, UK.
More informationTracking of Extended Object or Target Group Using Random Matrix Part I: New Model and Approach
Tracing of Extended Object or Target Group Using Random Matrix Part I: New Model and Approach Jian Lan Center for Information Engineering Science Research School of Electronics and Information Engineering
More informationQUADRATIC AND CONVEX MINIMAX CLASSIFICATION PROBLEMS
Journal of the Operations Research Societ of Japan 008, Vol. 51, No., 191-01 QUADRATIC AND CONVEX MINIMAX CLASSIFICATION PROBLEMS Tomonari Kitahara Shinji Mizuno Kazuhide Nakata Toko Institute of Technolog
More informationTwo Levels Data Fusion Filtering Algorithms of Multimode Compound Seeker
TELKOMNKA, Vol.11, No.11, November 13, pp. 674~68 e-ssn: 87-78X 674 Two Levels Data Fusion Filtering Algorithms of Multimode Compound Seeker Guodong Zhang*, Zhong Liu, Ke Li College of Electronic Engineering,
More informationAircraft Turbofan Engine Health Estimation Using Constrained Kalman Filtering
Cleveland State University EngagedScholarship@CSU Electrical Engineering & Computer Science Faculty Publications Electrical Engineering & Computer Science Department 6-16-2003 Aircraft Turbofan Engine
More informationParameterized Joint Densities with Gaussian and Gaussian Mixture Marginals
Parameterized Joint Densities with Gaussian and Gaussian Miture Marginals Feli Sawo, Dietrich Brunn, and Uwe D. Hanebeck Intelligent Sensor-Actuator-Sstems Laborator Institute of Computer Science and Engineering
More informationScattering Properties of Gas Molecules on a Water Adsorbed Surface
Scattering Properties of Gas Molecules on a Water Adsorbed Surface Hideki Takeuchi a, Koji Yamamoto b, and Toru Hakutake c a Department of Mechanical Engineering, Kochi National College of Technolog, Nankoku
More informationCombined Particle and Smooth Variable Structure Filtering for Nonlinear Estimation Problems
14th International Conference on Information Fusion Chicago, Illinois, USA, July 5-8, 2011 Combined Particle and Smooth Variable Structure Filtering for Nonlinear Estimation Problems S. Andrew Gadsden
More informationCPS 616 ITERATIVE IMPROVEMENTS 10-1
CPS 66 ITERATIVE IMPROVEMENTS 0 - APPROACH Algorithm design technique for solving optimization problems Start with a feasible solution Repeat the following step until no improvement can be found: change
More informationAutonomous Mobile Robot Design
Autonomous Mobile Robot Design Topic: Extended Kalman Filter Dr. Kostas Alexis (CSE) These slides relied on the lectures from C. Stachniss, J. Sturm and the book Probabilistic Robotics from Thurn et al.
More informationRECURSIVE OUTLIER-ROBUST FILTERING AND SMOOTHING FOR NONLINEAR SYSTEMS USING THE MULTIVARIATE STUDENT-T DISTRIBUTION
1 IEEE INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING, SEPT. 3 6, 1, SANTANDER, SPAIN RECURSIVE OUTLIER-ROBUST FILTERING AND SMOOTHING FOR NONLINEAR SYSTEMS USING THE MULTIVARIATE STUDENT-T
More informationTracking an Accelerated Target with a Nonlinear Constant Heading Model
Tracking an Accelerated Target with a Nonlinear Constant Heading Model Rong Yang, Gee Wah Ng DSO National Laboratories 20 Science Park Drive Singapore 118230 yrong@dsoorgsg ngeewah@dsoorgsg Abstract This
More information3.0 PROBABILITY, RANDOM VARIABLES AND RANDOM PROCESSES
3.0 PROBABILITY, RANDOM VARIABLES AND RANDOM PROCESSES 3.1 Introduction In this chapter we will review the concepts of probabilit, rom variables rom processes. We begin b reviewing some of the definitions
More informationDevelopment of an Extended Operational States Observer of a Power Assist Wheelchair
Development of an Extended Operational States Observer of a Power Assist Wheelchair Sehoon Oh Institute of Industrial Science Universit of Toko 4-6-, Komaba, Meguro, Toko, 53-855 Japan sehoon@horilab.iis.u-toko.ac.jp
More informationOptimal Gaussian Filtering for Polynomial Systems Applied to Association-free Multi-Target Tracking
4th International Conference on Information Fusion Chicago, Illinois, USA, July 5-8, Optimal Gaussian Filtering for Polynomial Systems Applied to Association-free Multi-Target Tracking Marcus Baum, Benjamin
More informationA comparative study of multiple-model algorithms for maneuvering target tracking
A comparative study of multiple-model algorithms for maneuvering target tracing Ryan R. Pitre Vesselin P. Jilov X. Rong Li Department of Electrical Engineering University of New Orleans New Orleans, LA
More informationGeneralized Linear Minimum Mean-Square Error Estimation
Generalized Linear Minimum Mean-Square Error Estimation Yu Liu and X. Rong Li Department of Electrical Engineering University of New Orleans New Orleans, LA 7148, U.S.A. Email: {lyu2, xli}@uno.edu Abstract
More informationInterspecific Segregation and Phase Transition in a Lattice Ecosystem with Intraspecific Competition
Interspecific Segregation and Phase Transition in a Lattice Ecosstem with Intraspecific Competition K. Tainaka a, M. Kushida a, Y. Ito a and J. Yoshimura a,b,c a Department of Sstems Engineering, Shizuoka
More informationManeuvering Target Tracking Method based on Unknown but Bounded Uncertainties
Preprints of the 8th IFAC World Congress Milano (Ital) ust 8 - Septemer, Maneuvering arget racing Method ased on Unnown ut Bounded Uncertainties Hodjat Rahmati*, Hamid Khaloozadeh**, Moosa Aati*** * Facult
More informationCh 3 Alg 2 Note Sheet.doc 3.1 Graphing Systems of Equations
Ch 3 Alg Note Sheet.doc 3.1 Graphing Sstems of Equations Sstems of Linear Equations A sstem of equations is a set of two or more equations that use the same variables. If the graph of each equation =.4
More informationA New Perspective and Extension of the Gaussian Filter
Robotics: Science and Sstems 2015 Rome, Ital, Jul 13-17, 2015 A New Perspective and Etension of the Gaussian Filter Manuel Wüthrich, Sebastian Trimpe, Daniel Kappler and Stefan Schaal Autonomous Motion
More informationPICONE S IDENTITY FOR A SYSTEM OF FIRST-ORDER NONLINEAR PARTIAL DIFFERENTIAL EQUATIONS
Electronic Journal of Differential Equations, Vol. 2013 (2013), No. 143, pp. 1 7. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu PICONE S IDENTITY
More informationOPTIMAL ESTIMATION of DYNAMIC SYSTEMS
CHAPMAN & HALL/CRC APPLIED MATHEMATICS -. AND NONLINEAR SCIENCE SERIES OPTIMAL ESTIMATION of DYNAMIC SYSTEMS John L Crassidis and John L. Junkins CHAPMAN & HALL/CRC A CRC Press Company Boca Raton London
More informationComparison of Fast ICA and Gradient Algorithms of Independent Component Analysis for Separation of Speech Signals
K. Mohanaprasad et.al / International Journal of Engineering and echnolog (IJE) Comparison of Fast ICA and Gradient Algorithms of Independent Component Analsis for Separation of Speech Signals K. Mohanaprasad
More informationBACKWARD FOKKER-PLANCK EQUATION FOR DETERMINATION OF MODEL PREDICTABILITY WITH UNCERTAIN INITIAL ERRORS
BACKWARD FOKKER-PLANCK EQUATION FOR DETERMINATION OF MODEL PREDICTABILITY WITH UNCERTAIN INITIAL ERRORS. INTRODUCTION It is widel recognized that uncertaint in atmospheric and oceanic models can be traced
More informationRecursive Noise Adaptive Kalman Filtering by Variational Bayesian Approximations
PREPRINT 1 Recursive Noise Adaptive Kalman Filtering by Variational Bayesian Approximations Simo Särä, Member, IEEE and Aapo Nummenmaa Abstract This article considers the application of variational Bayesian
More informationIntroduction to Unscented Kalman Filter
Introduction to Unscented Kalman Filter 1 Introdution In many scientific fields, we use certain models to describe the dynamics of system, such as mobile robot, vision tracking and so on. The word dynamics
More informationAn Accurate Incremental Principal Component Analysis method with capacity of update and downdate
0 International Conference on Computer Science and Information echnolog (ICCSI 0) IPCSI vol. 5 (0) (0) IACSI Press, Singapore DOI: 0.7763/IPCSI.0.V5. An Accurate Incremental Principal Component Analsis
More informationRiccati difference equations to non linear extended Kalman filter constraints
International Journal of Scientific & Engineering Research Volume 3, Issue 12, December-2012 1 Riccati difference equations to non linear extended Kalman filter constraints Abstract Elizabeth.S 1 & Jothilakshmi.R
More informationA New Perspective and Extension of the Gaussian Filter
A New Perspective and Etension of the Gaussian Filter The International Journal of Robotics Research 35(14):1 20 c The Author(s) 2016 Reprints and permission: sagepub.co.uk/journalspermissions.nav DOI:
More informationExperimental Uncertainty Review. Abstract. References. Measurement Uncertainties and Uncertainty Propagation
Experimental Uncertaint Review Abstract This is intended as a brief summar of the basic elements of uncertaint analsis, and a hand reference for laborator use. It provides some elementar "rules-of-thumb"
More informationEKF/UKF Maneuvering Target Tracking using Coordinated Turn Models with Polar/Cartesian Velocity
EKF/UKF Maneuvering Target Tracking using Coordinated Turn Models with Polar/Cartesian Velocity Michael Roth, Gustaf Hendeby, and Fredrik Gustafsson Dept. Electrical Engineering, Linköping University,
More informationRANGE CONTROL MPC APPROACH FOR TWO-DIMENSIONAL SYSTEM 1
RANGE CONTROL MPC APPROACH FOR TWO-DIMENSIONAL SYSTEM Jirka Roubal Vladimír Havlena Department of Control Engineering, Facult of Electrical Engineering, Czech Technical Universit in Prague Karlovo náměstí
More informationExtended Object and Group Tracking with Elliptic Random Hypersurface Models
Extended Object and Group Tracing with Elliptic Random Hypersurface Models Marcus Baum Benjamin Noac and Uwe D. Hanebec Intelligent Sensor-Actuator-Systems Laboratory ISAS Institute for Anthropomatics
More informationIEEE copyright notice
IEEE copright notice Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for
More informationDuality, Geometry, and Support Vector Regression
ualit, Geometr, and Support Vector Regression Jinbo Bi and Kristin P. Bennett epartment of Mathematical Sciences Rensselaer Poltechnic Institute Tro, NY 80 bij@rpi.edu, bennek@rpi.edu Abstract We develop
More informationSquare Estimation by Matrix Inverse Lemma 1
Proof of Two Conclusions Associated Linear Minimum Mean Square Estimation b Matri Inverse Lemma 1 Jianping Zheng State e Lab of IS, Xidian Universit, Xi an, 7171, P. R. China jpzheng@idian.edu.cn Ma 6,
More informationQuadrature Filters for Maneuvering Target Tracking
Quadrature Filters for Maneuvering Target Tracking Abhinoy Kumar Singh Department of Electrical Engineering, Indian Institute of Technology Patna, Bihar 813, India, Email: abhinoy@iitpacin Shovan Bhaumik
More informationStability Analysis for Linear Systems under State Constraints
Stabilit Analsis for Linear Sstems under State Constraints Haijun Fang Abstract This paper revisits the problem of stabilit analsis for linear sstems under state constraints New and less conservative sufficient
More informationNew Expression for the Functional Transformation of the Vector Cramér-Rao Lower Bound
New Expression for the Functional Transformation of the Vector Cramér-Rao Lower Bound Ali A. Nasir, Hani Mehrpouan, and Rodne A. Kenned, Research School of Information Sciences and Engineering, the Australian
More informationw u P z K Figure 1: Model of the closed loop control sstem Model In this section the model for the control problems will be given. It is static, multi
Optimal Linear Static Control with Moment and Yield Objectives Anders Hansson, Stephen Bod, Lieven Vandenberghe and Miguel Sousa Lobo Information Sstems Laborator Stanford Universit, Stanford, CA 94305{4055,
More informationA New Closed-loop Identification Method of a Hammerstein-type System with a Pure Time Delay
A New Closed-loop Identification Method of a Hammerstein-tpe Sstem with a Pure Time Dela Edin Drljević a, Branislava Peruničić b, Željko Jurić c, Member of IEEE Abstract A procedure for the closed-loop
More informationAn optimization approach to resolve the competing aims of active fault detection and control
211 5th IEEE Conference on Decision and Control and European Control Conference CDC-ECC) Orlando, FL, USA, December 12-15, 211 An optimization approach to resolve the competing aims of active fault detection
More informationLeast-Squares Independence Test
IEICE Transactions on Information and Sstems, vol.e94-d, no.6, pp.333-336,. Least-Squares Independence Test Masashi Sugiama Toko Institute of Technolog, Japan PRESTO, Japan Science and Technolog Agenc,
More informationOn Continuous-Discrete Cubature Kalman Filtering
On Continuous-Discrete Cubature Kalman Filtering Simo Särkkä Arno Solin Aalto University, P.O. Box 12200. FI-00076 AALTO, Finland. (Tel: +358 50 512 4393; e-mail: simo.sarkka@aalto.fi) Aalto University,
More informationNonlinear and/or Non-normal Filtering. Jesús Fernández-Villaverde University of Pennsylvania
Nonlinear and/or Non-normal Filtering Jesús Fernández-Villaverde University of Pennsylvania 1 Motivation Nonlinear and/or non-gaussian filtering, smoothing, and forecasting (NLGF) problems are pervasive
More informationNON-LINEAR NOISE ADAPTIVE KALMAN FILTERING VIA VARIATIONAL BAYES
2013 IEEE INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING NON-LINEAR NOISE ADAPTIVE KALMAN FILTERING VIA VARIATIONAL BAYES Simo Särä Aalto University, 02150 Espoo, Finland Jouni Hartiainen
More informationThe Unscented Particle Filter
The Unscented Particle Filter Rudolph van der Merwe (OGI) Nando de Freitas (UC Bereley) Arnaud Doucet (Cambridge University) Eric Wan (OGI) Outline Optimal Estimation & Filtering Optimal Recursive Bayesian
More informationAdaptive Unscented Kalman Filter with Multiple Fading Factors for Pico Satellite Attitude Estimation
Adaptive Unscented Kalman Filter with Multiple Fading Factors for Pico Satellite Attitude Estimation Halil Ersin Söken and Chingiz Hajiyev Aeronautics and Astronautics Faculty Istanbul Technical University
More informationCity, University of London Institutional Repository
City Research Online City, University of London Institutional Repository Citation: Zhao, S., Shmaliy, Y. S., Khan, S. & Liu, F. (2015. Improving state estimates over finite data using optimal FIR filtering
More informationIn-Orbit Magnetometer Calibration for a Spinning Satellite
In-Orit Magnetometer Caliration for a Spinning Satellite Halil Ersin Söen and Shin-ichiro Saai Japan Aerospace Exploration Agenc (JAXA), Sagamihara, 5-50 Japan Magnetometers are commonl used attitude sensors
More informationOrdinal One-Switch Utility Functions
Ordinal One-Switch Utilit Functions Ali E. Abbas Universit of Southern California, Los Angeles, California 90089, aliabbas@usc.edu David E. Bell Harvard Business School, Boston, Massachusetts 0163, dbell@hbs.edu
More informationUsing the Kalman Filter to Estimate the State of a Maneuvering Aircraft
1 Using the Kalman Filter to Estimate the State of a Maneuvering Aircraft K. Meier and A. Desai Abstract Using sensors that only measure the bearing angle and range of an aircraft, a Kalman filter is implemented
More informationEstimation of the Vehicle Sideslip Angle by Means of the State Dependent Riccati Equation Technique
Proceedings o the World Congress on Engineering 7 Vol II WCE 7, Jul 5-7, 7, London, U.K. Estimation o the Vehicle Sideslip Angle b Means o the State Dependent Riccati Equation Technique S. Strano, M. Terzo
More informationA Comparison of Multiple-Model Target Tracking Algorithms
University of New Orleans ScholarWors@UNO University of New Orleans heses and Dissertations Dissertations and heses 1-17-4 A Comparison of Multiple-Model arget racing Algorithms Ryan Pitre University of
More informationApplication of Unscented Transformation for Nonlinear State Smoothing
Application of Unscented Transformation for Nonlinear State Smoothing WANG Xiao-Xu 1 PAN Quan 1 LIANG Yan 1 ZHAO Chun-Hui 1 Abstract Motivated by the well-known fact that the state estimate of a smoother
More informationWeek 3 September 5-7.
MA322 Weekl topics and quiz preparations Week 3 September 5-7. Topics These are alread partl covered in lectures. We collect the details for convenience.. Solutions of homogeneous equations AX =. 2. Using
More informationCONSTRAINED STATE ESTIMATION IN PARTICLE FILTERS
COSTRAIED STATE ESTIMATIO I PARTICLE FILTERS Bradley Ebinger, idhal Bouaynaya and Robi Polikar * Roman Shterenberg Rowan University Department of Electrical and Computer Engineering Glassboro, J University
More informationA student's t filter for heavy tailed process and measurement noise
A student's t filter for heavy tailed process and measurement noise Michael Roth, Emre Ozan and Fredri Gustafsson Linöping University Post Print N.B.: When citing this wor, cite the original article. Original
More informationDistributed estimation in sensor networks
in sensor networks A. Benavoli Dpt. di Sistemi e Informatica Università di Firenze, Italy. e-mail: benavoli@dsi.unifi.it Outline 1 An introduction to 2 3 An introduction to An introduction to In recent
More information1 Computing with constraints
Notes for 2017-04-26 1 Computing with constraints Recall that our basic problem is minimize φ(x) s.t. x Ω where the feasible set Ω is defined by equality and inequality conditions Ω = {x R n : c i (x)
More informationMini-Course 07 Kalman Particle Filters. Henrique Massard da Fonseca Cesar Cunha Pacheco Wellington Bettencurte Julio Dutra
Mini-Course 07 Kalman Particle Filters Henrique Massard da Fonseca Cesar Cunha Pacheco Wellington Bettencurte Julio Dutra Agenda State Estimation Problems & Kalman Filter Henrique Massard Steady State
More informationMYSTERY BEHIND SIMILARITY MEASURES MSE AND SSIM. Gintautas Palubinskas
MYSTEY BEHID SIMILAITY MEASUES MSE AD SSIM Gintautas Palubinskas German Aerospace Center DL, emote Sensing Technolog Institute Oberpfaffenhofen, 834 Wessling, German ABSTACT Similarit or distance measures
More informationFractal dimension of the controlled Julia sets of the output duopoly competing evolution model
Available online at www.isr-publications.com/jmcs J. Math. Computer Sci. 1 (1) 1 71 Research Article Fractal dimension of the controlled Julia sets of the output duopol competing evolution model Zhaoqing
More informationA STUDY ON THE STATE ESTIMATION OF NONLINEAR ELECTRIC CIRCUITS BY UNSCENTED KALMAN FILTER
A STUDY ON THE STATE ESTIMATION OF NONLINEAR ELECTRIC CIRCUITS BY UNSCENTED KALMAN FILTER Esra SAATCI Aydın AKAN 2 e-mail: esra.saatci@iku.edu.tr e-mail: akan@istanbul.edu.tr Department of Electronic Eng.,
More informationState Estimation with Unconventional and Networked Measurements
University of New Orleans ScholarWors@UNO University of New Orleans Theses and Dissertations Dissertations and Theses 5-14-2010 State Estimation with Unconventional and Networed Measurements Zhansheng
More informationPOLYNOMIAL EXPANSION OF THE PROBABILITY DENSITY FUNCTION ABOUT GAUSSIAN MIXTURES
POLYNOMIAL EXPANSION OF THE PROBABILITY DENSITY FUNCTION ABOUT GAUSSIAN MIXTURES Charles C. Cavalcante, J. C. M. Mota and J. M. T. Romano GTEL/DETI/CT/UFC C.P. 65, Campus do Pici, CEP: 6.455-76 Fortaleza-CE,
More informationDiscontinuous Galerkin method for a class of elliptic multi-scale problems
INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN FLUIDS Int. J. Numer. Meth. Fluids 000; 00: 6 [Version: 00/09/8 v.0] Discontinuous Galerkin method for a class of elliptic multi-scale problems Ling Yuan
More informationPrediction of ESTSP Competition Time Series by Unscented Kalman Filter and RTS Smoother
Prediction of ESTSP Competition Time Series by Unscented Kalman Filter and RTS Smoother Simo Särkkä, Aki Vehtari and Jouko Lampinen Helsinki University of Technology Department of Electrical and Communications
More informationDATA FUSION III: Estimation Theory
Data Fusion III Estimation Theor March 16, 006 DATA FUSION III: Estimation Theor Date: March 16, 006 Time: 5:00 7:30 PM Location: B-300--3 (AAR-400) (Main Building, nd floor, near freight elevators) Instructor:
More informationVibration of Plate on Foundation with Four Edges Free by Finite Cosine Integral Transform Method
854 Vibration of Plate on Foundation with Four Edges Free b Finite Cosine Integral Transform Method Abstract The analtical solutions for the natural frequencies and mode shapes of the rectangular plate
More informationA new unscented Kalman filter with higher order moment-matching
A new unscented Kalman filter with higher order moment-matching KSENIA PONOMAREVA, PARESH DATE AND ZIDONG WANG Department of Mathematical Sciences, Brunel University, Uxbridge, UB8 3PH, UK. Abstract This
More informationarxiv: v1 [cs.sy] 8 May 2018
Tracing the Orientation and Axes Lengths of an Elliptical Extended Object Shishan Yang a, Marcus Baum a a Institute of Computer Science University of Göttingen, Germany arxiv:185.376v1 [cs.sy] 8 May 18
More informationUNCORRECTED SAMPLE PAGES. 3Quadratics. Chapter 3. Objectives
Chapter 3 3Quadratics Objectives To recognise and sketch the graphs of quadratic polnomials. To find the ke features of the graph of a quadratic polnomial: ais intercepts, turning point and ais of smmetr.
More informationEstimation Of Linearised Fluid Film Coefficients In A Rotor Bearing System Subjected To Random Excitation
Estimation Of Linearised Fluid Film Coefficients In A Rotor Bearing Sstem Subjected To Random Ecitation Arshad. Khan and Ahmad A. Khan Department of Mechanical Engineering Z.. College of Engineering &
More informationMoving Horizon Estimation with Decimated Observations
Moving Horizon Estimation with Decimated Observations Rui F. Barreiro A. Pedro Aguiar João M. Lemos Institute for Systems and Robotics (ISR) Instituto Superior Tecnico (IST), Lisbon, Portugal INESC-ID/IST,
More informationChapter 4 Analytic Trigonometry
Analtic Trigonometr Chapter Analtic Trigonometr Inverse Trigonometric Functions The trigonometric functions act as an operator on the variable (angle, resulting in an output value Suppose this process
More information