In this thesis, we investigate a decentralized data fusion system with improved particle filters for target tracking. In many application areas, it becomes essential to use nonlinear and non-Gaussian elements to accurately model the underlying dynamics of a physical system. Particle filters have a great potential for solving highly nonlinear and non-Gaussian estimation problems, in which the traditional Kalman filter and extended Kalman filter may generally fail. To improve the tracking performance of particle filters, initialization of the particles is studied. We construct an initial state distribution by using least square estimation. In addition, to enhance the tracking capability of particle filters, representation of target velocity by another set of particles is considered. We include another layer of particle filter inside the original particle filter for updating the velocity. In our proposed architecture, we assume that each sensor node contain a particle filter and there is no fusion center in the sensor network. Approximated a posteriori distribution at the sensor is obtained by using the local particle filters with the Gaussian mixture model (GMM), so that more compact representations of the distribution for transmission can be obtained. To achieve information sharing and integration, the GMM-covariance intersection algorithm is used in formulating the decentralized fusion solutions. Simulation results are presented to illustrate that the performance of the improved particle filter is better than standard particle filter. In addition, simulation results of target tracking in the sensor system with three sensor nodes are given to show the effectiveness and superiority of the proposed architecture.

This thesis details the development of a custom autopilot system designed specifically for multi-agent robotic missions. The project was motivated by the need for a flexible autopilot system architecture that could be easily adapted to a variety of future multi-vehicle experiments. The development efforts can be split into three categories: algorithm and software development, hardware development, and testing and integration. Over 12,000 lines of C++ code were written in this project, resulting in custom flight and ground control software. The flight software was designed to run on a Gumstix Overo Fire(STORM) computer on module (COM) using a Linux Angstrom operating system. The flight software was designed to support the onboard GN&C algorithms. The ground control station and its graphical user interface were developed in the Qt C++ framework. The ground control software has been proven to operate safely during multi-vehicle tests, and will be an asset in future work. Two TSH GAUI 500X quad-rotors and one Gears Educational Systems SMP rover were integrated into an autonomous swarm. Each vehicle used the Gumstix Overo COM. The C-DUS Pilot board was designed as a custom interface circuit board for the Overo COM and its expansion board, the Gumstix Pinto-TH. While the built-in WiFi capability of the Overo COM served as a communication link to a central wireless router, the C-DUS Pilot board allowed for the compact and reliable integration of sensors and actuators. The sensors used in this project were limited to accelerometers, gyroscopes, magnetometers, and GPS. All of the components underwent extensive testing. A series of ground and flight tests were conducted to safely and gradually prove system capabilities. The work presented in this thesis culminated with a successful three-vehicle autonomous demonstration comprised of two quad-rotors executing a standoff tracking trajectory around a moving rover, while simultaneously performing GPS-based collision avoidance. / text

No description available.

Tracking multiple targets in a high cluttered environment where multiple receivers are used is a challenging task due to the high level of false alarms and uncertainty in the track hypothesis. The multi-static active sonar scenario is an example for such systems where multiple source-receiver combinations are deployed. Due to the nature of the underwater environment and sound propagation characteristics, tracking targets in the underwater environment becomes a complex operation. Conventional tracking approaches (such as the Kalman and particle filter) require a predetermined kinematic model of the target. Moreover, tracking an unknown and changing number of targets within a certain search area requires complex mathematical association filters to identify the number of targets and associate measurements to different target tracks. As the number of false detections increases, the computational complexity of conventional tracking system grows introducing further challenges for real-time target tracking situations. The methodology presented in this thesis provides a rapid and reliable tracking system capable of tracking multiple targets without depending on a kinematic model of the target movement. In this algorithm, Self Organizing Maps, Dynamic Programming and the Hough transform are combined to produce tracks of possible targets’ paths and estimate of targets’ locations. Evaluation of the performance of the tracking algorithm is performed using three types of simulations and a set of real data obtained from a sea trial. This research documents the results of experimental testing and analysis of the tracking system. / Thesis (Master, Electrical & Computer Engineering) -- Queen's University, 2009-08-07 13:21:06.869

In this thesis we present LaSLAT, a sensor network algorithm thatsimultaneously localizes sensors, calibrates sensing hardware, andtracks unconstrained moving targets using only range measurementsbetween the sensors and the target. LaSLAT is based on a Bayesian filter, which updates a probabilitydistribution over the quantities of interest as measurementsarrive. The algorithm is distributable, and requires only a constantamount of space with respect to the number of measurementsincorporated. LaSLAT is easy to adapt to new types of hardware and newphysical environments due to its use of intuitive probabilitydistributions: one adaptation demonstrated in this thesis uses amixture measurement model to detect and compensate for bad acousticrange measurements due to echoes.We also present results from a centralized Java implementation ofLaSLAT on both two- and three-dimensional sensor networks in whichranges are obtained using the Cricket ranging system. LaSLAT is ableto localize sensors to within several centimeters of their groundtruth positions while recovering a range measurement bias for eachsensor and the complete trajectory of the mobile.

abstract: The tracking of multiple targets becomes more challenging in complex environments due to the additional degrees of nonlinearity in the measurement model. In urban terrain, for example, there are multiple reflection path measurements that need to be exploited since line-of-sight observations are not always available. Multiple target tracking in urban terrain environments is traditionally implemented using sequential Monte Carlo filtering algorithms and data association techniques. However, data association techniques can be computationally intensive and require very strict conditions for efficient performance. This thesis investigates the probability hypothesis density (PHD) method for tracking multiple targets in urban environments. The PHD is based on the theory of random finite sets and it is implemented using the particle filter. Unlike data association methods, it can be used to estimate the number of targets as well as their corresponding tracks. A modified maximum-likelihood version of the PHD (MPHD) is proposed to automatically and adaptively estimate the measurement types available at each time step. Specifically, the MPHD allows measurement-to-nonlinearity associations such that the best matched measurement can be used at each time step, resulting in improved radar coverage and scene visibility. Numerical simulations demonstrate the effectiveness of the MPHD in improving tracking performance, both for tracking multiple targets and targets in clutter. / Dissertation/Thesis / M.S. Electrical Engineering 2011

Many multiple target tracking algorithms operate in the local frame of the sensor and have difficulty with track reallocation when targets move in and out of the sensor field of view. This poses a problem when an unmanned aerial vehicle (UAV) is tracking multiple ground targets on a road network larger than its field of view. We propose a Rao-Blackwellized Particle Filter (RBPF) to maintain individual target tracks and to perform probabilistic data association when the targets are constrained to a road network. This is particularly useful when a target leaves then re-enters the UAV’s field of view. The RBPF is structured as a particle filter of particle filters. The top level filter handles data association and each of its particles maintains a bank of particle filters to handle target tracking. The tracking particle filters incorporate both positive and negative information when a measurement is received. We implement two path planning controllers, exhaustive receding horizon control (ERHC) and a neural net trained with deep reinforcement learning (Deep-RL), and compare their ability to improve the certainty for multiple target location estimates. The controllers prioritize paths that reduce each target’s entropy. While the ERHC achieved optimal stead-state estimates the DeepRL controller identified more efficient sweeping search patterns when there is limited information regarding target locations. The neural net achieves O(1) computational complexity during decision making but must first be trained on a given map. In addition, we provide a theorem that calculates the lower-bound for the average-entropy of the RBPF. Particle Filter entropy is used as a unit of measurement as it gives a way of accurately comparing the precision of complex multi-modal estimates. This gives a reliable way of establishing the resources needed to accomplish mission objectives as well as providing a reliable method of determining the effectiveness of different multi-agent path planners. Finally we outline results both in simulation and hardware. In simulation we obtained the results for our different path planners over 2000 Monte Carlo runs and show how the different path planners compare and measure up to the lower-bound of average-entropy. The results from a hardware test provide evidence that the ideas presented in this thesis hold true in an end-to-end solution.

Many multiple target tracking algorithms operate in the local frame of the sensor and have difficulty with track reallocation when targets move in and out of the sensor field of view. This poses a problem when an unmanned aerial vehicle (UAV) is tracking multiple ground targets on a road network larger than its field of view. We propose a Rao-Blackwellized Particle Filter (RBPF) to maintain individual target tracks and to perform probabilistic data association when the targets are constrained to a road network. This is particularly useful when a target leaves then re-enters the UAV's field of view. The RBPF is structured as a particle filter of particle filters. The top level filter handles data association and each of its particles maintains a bank of particle filters to handle target tracking. The tracking particle filters incorporate both positive and negative information when a measurement is received. We implement two path planning controllers, exhaustive receding horizon control (ERHC) and a neural net trained with deep reinforcement learning (Deep-RL), and compare their ability to improve the certainty for multiple target location estimates. The controllers prioritize paths that reduce each target's entropy. While the ERHC achieved optimal stead-state estimates the Deep-RL controller identified more efficient sweeping search patterns when there is limited information regarding target locations. The neural net achieves O(1) computational complexity during decision making but must first be trained on a given map. In addition, we provide a theorem that calculates the lower-bound for the average-entropy of the RBPF. Particle Filter entropy is used as a unit of measurement as it gives a way of accurately comparing the precision of complex multi-modal estimates. This gives a reliable way of establishing the resources needed to accomplish mission objectives as well as providing a reliable method of determining the effectiveness of different multi-agent path planners. Finally we outline results both in simulation and hardware. In simulation we obtained the results for our different path planners over 2000 Monte Carlo runs and show how the different path planners compare and measure up to the lower-bound of average-entropy. The results from a hardware test provide evidence that the ideas presented in this thesis hold true in an end-to-end solution.

<p> In this thesis new joint target tracking and classification techniques for aspect-dependent measurements are developed. Joint target tracking and classification methods can result in better tracking and classification performance than those treating these as two separate problems. Significant improvement in state estimation and classification performance can be achieved by exchanging useful information between the tracker and the classifier. Target classification in many target tracking algorithms is not typically done by taking into consideration the target-to-sensor orientation. However, the feature information extracted from the signal that originated from the target is generally a strong function of the target-to-sensor orientation. Since sensor returns are sensitive to this orientation, classification from a single sensor may not give exact target classes. Better classification results can be obtained by fusing feature measurements from multiple views of a target. In multitarget scenarios, handling the classification becomes more challenging due to the identifying the feature information corresponding to a target. That is, it is difficult to identify the origin of measurements. In this case, feature measurement origin ambiguities can be eliminated by integrating the classifier into multiframe data association. This technique reduces the ambiguity in feature measurements while improving track purity. </p> <p> A closed form expression for multiaspect target classification is not feasible. Then, training based statistical modeling can be used to model the unknown feature measurements of a target. In this thesis, the Observable Operator Model (OOM), a better alternative to the Hidden Markov Model (HMM), is used to capture unknown feature distribution of each target and thus can be used as a classifier. The proposed OOM based classification technique incorporates target-to-sensor orientation with a sequence of feature information from multiple sensors. Further, the multi-aspect classifier can be modeled using the OOM to handle unknown target orientation. The target orientation estimation using OOM can also be used to find improved estimates of the states of highly maneuverable targets with noisy kinematic measurements. One limiting factor in obtaining accurate estimates of highly maneuvering target states is the high level of uncertainty in velocity and acceleration components. The target orientation information is helpful in alleviating this problem to accurately determine the velocity and acceleration components. </p> <p> Various simulation studies based on two-dimensional scenarios are presented in this thesis to demonstrate the merits of the proposed joint target tracking and classification algorithms that use aspect-dependent feature measurements.</p> / Thesis / Doctor of Philosophy (PhD)

With the recent advances in sensor technology and resulting sensor resolution, conven- tional point-based target tracking algorithms are becoming insufficient, particularly in application domains such as autonomous vehicles, visual tracking and surveillance using high resolution sensors. This has renewed the interest in extended target (ET) tracking, which aims to track not only the centroid of a target, but also its shape and size over time. This thesis addresses three of the most challenging problems in the domain of ET tracking applications. The first investigated challenge is the need for an accu- rate shape and centre estimate for the ET object with an arbitrary unknown star- convex shape in presence of non-Gaussian noise. The proposed method is based on a Student’s-t process regression algorithm which is defined in a recursive framework to be applicable for on-line tracking problems. The second problem tries to relax any constraints, including the star-convex con- straint, that is imposed on the shape of the ET object during the course of estimation by defining a novel Random Polytopes shape descriptor. Also, the proposed solution introduces a method to mitigate the troubles caused as a result of self-occlusion in ET tracking applications which its ignorance may cause catastrophic divergence in the ET state estimate.Finally, a framework for tracking multiple ET objects in the presence of clutter and occlusion is studied and a solution is proposed. The proposed method can estimate the centre and shape of the ET objects in a realistically scenario with the self- and mutual-occlusion challenges being considered. The proposed approach defines a time varying state-dependent probability of detection for each ET that enables the track to prolong even under adverse conditions caused due to mutual-occlusion. Plus, the proposed algorithm uses set-membership uncertainty models to bound the association and target shape uncertainties of occluded ET, to obtain more accurate state and shape estimates of an ET object. The performance of the proposed methods are quantified on realistically simulated scenarios with self- and mutual-occlusions and their results are compared against existing state-of-the-art methods for ET tracking applications. / Thesis / Doctor of Philosophy (PhD)