Browsing by Subject "sensor planning"

This thesis addresses the problem of information-driven sensor path planning for the purpose of target detection, measurement, and classification using non-holonomic mobile sensor agents (MSAs). Each MSA is equipped with two types of sensors. One is the measuring sensor with small FOV, while the other is the detecting sensor with large FOV. The measuring sensor could be ground penetrating radar (GPR), while the detecting sensor can be infrared radar (IR). The classification of a target can be reduced to the problem of estimating one or more random variables associated with this target from partial or imperfect measurements from sensorscite{stengel}, and can be represented by a probability mass function (PMF). Previous work shows the performance of MSAs can be greatly improved by planning their motion and control laws based on their sensing objectives. Because of the stochastic nature of sensing objective, the expected measurement benefit of a target, i.e, the information value, is defined as the expected entropy reduction of its classification PMF before the next measurement is taken of this target. The information value of targets is combined with other robot motion planning methods to address the sensor planning problem.

By definition, the entropy reduction can be represented by conditional mutual information of PMF given a measurement. MSAs are deployed in an obstacle-populated environment, and must avoid collisions with obstacles, as well as, in some cases, targets.

This thesis first presents a modified rapidly-exploring random trees (RRTs) approach with a novel milestone sampling method. The sampling function for RRTs takes into account the information value of targets, and sensor measurements of obstacle locations, as well as MSAs' configurations (e.g., position and orientation) and velocities to generate new milestones for expanding the trees online. By considering the information value, the sample function favors expansions toward targets with higher expected measurement benefit. After sampling, the MSAs navigate to the selected milestones based on the critic introduced later and take measurements of targets within their FOVs. Then, the thesis introduces an information potential method (IPM) approach which combined information values of targets with the potential functions. Targets with high information value have larger influence distance and tend to have high probability to be measured by the MSAs. Additionally, this information potential field is utilized to generate the milestones in a local probabilistic roadmap method to help MSAs escape their local minima.

The proposed two methods are applied to a landmine classification problem. It is assumed that geometries and locations of partial obstacles and targets are available as prior information, as well as previous measurements on targets concerning their classification. The experiments show that paths of MSAs using the modified RRTs and IPM take advantages of the information value by favoring targets with high information value. Furthermore, the results show that the IPM

outperforms other approaches such as the modified RRTs with information value and classical potential field method that does not take target information value into account.

Bayesian nonparametric models, such as the Gaussian process and the Dirichlet process, have been extensively applied for target kinematics modeling in various applications including environmental monitoring, traffic planning, endangered species tracking, dynamic scene analysis, autonomous robot navigation, and human motion modeling. As shown by these successful applications, Bayesian nonparametric models are able to adjust their complexities adaptively from data as necessary, and are resistant to overfitting or underfitting. However, most existing works assume that the sensor measurements used to learn the Bayesian nonparametric target kinematics models are obtained a priori or that the target kinematics can be measured by the sensor at any given time throughout the task. Little work has been done for controlling the sensor with bounded field of view to obtain measurements of mobile targets that are most informative for reducing the uncertainty of the Bayesian nonparametric models. To present the systematic sensor planning approach to leaning Bayesian nonparametric models, the Gaussian process target kinematics model is introduced at first, which is capable of describing time-invariant spatial phenomena, such as ocean currents, temperature distributions and wind velocity fields. The Dirichlet process-Gaussian process target kinematics model is subsequently discussed for modeling mixture of mobile targets, such as pedestrian motion patterns.

Novel information theoretic functions are developed for these introduced Bayesian nonparametric target kinematics models to represent the expected utility of measurements as a function of sensor control inputs and random environmental variables. A Gaussian process expected Kullback Leibler divergence is developed as the expectation of the KL divergence between the current (prior) and posterior Gaussian process target kinematics models with respect to the future measurements. Then, this approach is extended to develop a new information value function that can be used to estimate target kinematics described by a Dirichlet process-Gaussian process mixture model. A theorem is proposed that shows the novel information theoretic functions are bounded. Based on this theorem, efficient estimators of the new information theoretic functions are designed, which are proved to be unbiased with the variance of the resultant approximation error decreasing linearly as the number of samples increases. Computational complexities for optimizing the novel information theoretic functions under sensor dynamics constraints are studied, and are proved to be NP-hard. A cumulative lower bound is then proposed to reduce the computational complexity to polynomial time.

Three sensor planning algorithms are developed according to the assumptions on the target kinematics and the sensor dynamics. For problems where the control space of the sensor is discrete, a greedy algorithm is proposed. The efficiency of the greedy algorithm is demonstrated by a numerical experiment with data of ocean currents obtained by moored buoys. A sweep line algorithm is developed for applications where the sensor control space is continuous and unconstrained. Synthetic simulations as well as physical experiments with ground robots and a surveillance camera are conducted to evaluate the performance of the sweep line algorithm. Moreover, a lexicographic algorithm is designed based on the cumulative lower bound of the novel information theoretic functions, for the scenario where the sensor dynamics are constrained. Numerical experiments with real data collected from indoor pedestrians by a commercial pan-tilt camera are performed to examine the lexicographic algorithm. Results from both the numerical simulations and the physical experiments show that the three sensor planning algorithms proposed in this dissertation based on the novel information theoretic functions are superior at learning the target kinematics with

little or no prior knowledge