Download presentation
Presentation is loading. Please wait.
Published byLuke Jordan Modified over 9 years ago
1
Rickard Karlsson IEEE Aerospace Conf 2007 Target Tracking Performance Evaluation A General Software Environment for Filtering Rickard Karlsson Gustaf Hendeby Automatic Control Linköping University, SWEDEN rickard@isy.liu.se
2
Rickard Karlsson IEEE Aerospace Conf 2007 Motivating Example Range-Only Measurements Two Sensors with range uncertainties Performance? General Software for filtering
3
Rickard Karlsson IEEE Aerospace Conf 2007 Motivating Example: General Filtering Software Surface/UW navigation Robotics Bearings-Only Tracking Track Before Detect Performance usually using RMSE Applications
4
Rickard Karlsson IEEE Aerospace Conf 2007 Outline Nonlinear filtering using particle filters Performace measure for nonlinear filtering Kullback-Divergence vs RMSE General Filtering Software Object oriented design Design Patterns Examples
5
Rickard Karlsson IEEE Aerospace Conf 2007 Filtering STATE SPACE MODEL Process noise Measurement noise PROBABILISTIC DESCRIPTION
6
Rickard Karlsson IEEE Aerospace Conf 2007 Bayesian Recursions: Probability Density Function (pdf) M.U. T.U.
7
Rickard Karlsson IEEE Aerospace Conf 2007 Filter Evaluation: Mean Square Error (MSE) Mean square error (MSE) Standard performance measure Approximates the estimation error covariance Bounded by the Cramér-Rao Lower Bound (CRLB) Ignores higher-order moments! Compare the true trajectory with the estimated!!! What can we do instead?
8
Rickard Karlsson IEEE Aerospace Conf 2007 Kullback-Leibler Information
9
Rickard Karlsson IEEE Aerospace Conf 2007 Filter Evaluation: Kullback Divergence (KD) Kullback Divergence (KD) Compares the distance between two distributions Captures all moments of the distributions True PDF from a grid-based method True PDF from PF, compare sub-optimal filters Smoothing kernel needed for implementation Compare the true PDF with the estimated PDF.
10
Rickard Karlsson IEEE Aerospace Conf 2007 Generalized Gaussian Generalized Gaussion Distribution Kullback Divergence PDF
11
Rickard Karlsson IEEE Aerospace Conf 2007 Example 1: One-dimensional Nonlinear System Probability Density Function x Time
12
Rickard Karlsson IEEE Aerospace Conf 2007 Example 1: One-dimensional Nonlinear System Kullback Divergence RMSE KD for one realization comparing PF and EKF RMSE for 400 MC simulations
13
Rickard Karlsson IEEE Aerospace Conf 2007 Example 2: Range-Only Measurement Estimate target position from range-only measurements Nonlinear measurements but Gaussian noise Posterior distribution: bimodal Point Estimate: EKF vs PF the same, i.e. same RMSE
14
Rickard Karlsson IEEE Aerospace Conf 2007 Example 2: Simulation Results for Range-Only MSE KD No Difference! KD Indicates a Difference! EKF PF EKF PF
15
Rickard Karlsson IEEE Aerospace Conf 2007 Calculating the probability EKF PF&True Probability for target within the circle with radius R
16
Rickard Karlsson IEEE Aerospace Conf 2007 Example 3: Linear -- Non-Gaussian Noise MSE similar for both KF and PF! KL is better for PF, which is accounted for by multimodal target distribution which is closer to the truth
17
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Environment in C++ MATLAB Easy to use Weak typing Somewhat slow Object oriented (not really) C++ More complicated to use Fast Strong typing Object oriented Can be implemented ! F++: Fairly easy to use Just provide models f(x), h(x), etc Estimators: EKF, PF, IMM, UKF Open Source code available www.control.isy.liu.se/resources/f++ OOP & Design Patterns
18
Rickard Karlsson IEEE Aerospace Conf 2007 Object Oriented Programming (OOP) Inheritance Encapsulation Overloading
19
Rickard Karlsson IEEE Aerospace Conf 2007 Design Patterns – What is it? Smart Pointers Singletons Object factories … “Design patterns are general, programming language independent, conceptual high level solutions to common problems” Example:
20
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML
21
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML
22
Rickard Karlsson IEEE Aerospace Conf 2007 Class: Model
23
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML
24
Rickard Karlsson IEEE Aerospace Conf 2007 Class: Noise
25
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML
26
Rickard Karlsson IEEE Aerospace Conf 2007 Class: Estimator
27
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML Ex: Linear Gaussian system with KF and MATLAB support
28
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML Ex: Non-Linear Gaussian system with PF and MATLAB support
29
Rickard Karlsson IEEE Aerospace Conf 2007 Particle Filtering in Practice General model Common models for tracking and navigation
30
Rickard Karlsson IEEE Aerospace Conf 2007 F++ A General Filtering Framwork in C++ ModelNoiseEstimatorI/O LinModel MultiModel GenericModel Gauss SumNoise … EKF PF IMM UKF MPF MATLAB XML Ex: Linear Dynamics, Non-Linear Measurements Gaussian noise with PF and MATLAB support
31
Rickard Karlsson IEEE Aerospace Conf 2007 Code: Main Estimation Loop Time Update Meas. Update Estimate TU/MU/Est This works for any estimator! estimate u y filter
32
Rickard Karlsson IEEE Aerospace Conf 2007 Code: Main Estimation Loop Estimator Time Update Meas. Update Estimate This works for any estimator! estimate u y filter
33
Rickard Karlsson IEEE Aerospace Conf 2007 Code: Main Program INPUT MC-loop True/Meas Estimate OUTPUT
34
Rickard Karlsson IEEE Aerospace Conf 2007 Code: Main Program INPUT MC-loop True/Meas Estimate OUTPUT
35
Rickard Karlsson IEEE Aerospace Conf 2007 Code: Input (from MATLAB) Just type f(x), h(x),… Nonlinear? Matrices!!
36
Rickard Karlsson IEEE Aerospace Conf 2007 Example 4: Positioning using IMU and GPS MPF part of next release
37
Rickard Karlsson IEEE Aerospace Conf 2007 Summary Rickard Karlsson Automatic Control Linköping University, SWEDEN rickard@isy.liu.se www.control.isy.liu.se/~rickard Proposed KD as a performance measure General Filtering Software
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.