Amit S. Chhetri

Energy efficient target tracking in a sensor network using non-myopic sensor scheduling

We propose to use non-myopic sensor scheduling to minimize the energy usage in a sensor network while maintaining a desired squared-error tracking accuracy of a targets position estimate. The network comprises of Type A sensors that collect measurements, and Type B sensors that collect, process, and schedule measurements. The target is tracked using a particle filter; only Type B sensors hold the target belief and update it with measurements. Network energy consumption is primarily due to sensing and communicating belief and measurements between sensors. To schedule a sequence of M sensing actions, the Type B sensor holding the target belief computes the minimum energy sequence that satisfies the tracking accuracy constraint M steps in the future. Scheduling is implemented efficiently by precomputing an energy tree and using a uniform-cost search. The tracking accuracy for sensor scheduling is approximated by the posterior Cramer-Rao lower bound. Using Monte Carlo simulations, we demonstrate that non-myopic scheduling results in significantly lower energy usage than myopic scheduling while meeting the accuracy constraint.

Scheduling multiple sensors using particle filters in target tracking

A critical component of a multi-sensor system is sensor scheduling to optimize system performance under constraints (e.g. power, bandwidth, and computation). In this paper, we apply particle filter sequential Monte Carlo methods to implement multiple sensor scheduling for target tracking. Under the constraint that only one sensor can be used at each time step, we select a sequence of sensor uses to minimize the predicted mean-square error in the target state estimate; the predicted mean-square error is approximated using the particle filter in conjunction with an extended Kaiman filter approximation. Using Monte Carlo simulations, we demonstrate the improved performance of our scheduling approach over the non-scheduling case.

Nonmyopic sensor scheduling and its efficient implementation for target tracking applications

We propose two nonmyopic sensor scheduling algorithms for target tracking applications. We consider a scenario where a bearing-only sensor is constrained to move in a finite number of directions to track a target in a two-dimensional plane. Both algorithms provide the best sensor sequence by minimizing a predicted expected scheduler cost over a finite time-horizon. The first algorithm approximately computes the scheduler costs based on the predicted covariance matrix of the tracker error. The second algorithm uses the unscented transform in conjunction with a particle filter to approximate covariance-based costs or information-theoretic costs. We also propose the use of two branch-and-bound-based optimal pruning algorithms for efficient implementation of the scheduling algorithms. We design the first pruning algorithm by combining branch-and-bound with a breadth-first search and a greedy-search; the second pruning algorithm combines branch-and-bound with a uniform-cost search. Simulation results demonstrate the advantage of nonmyopic scheduling over myopic scheduling and the significant savings in computational and memory resources when using the pruning algorithms.

On the Use of Binary Programming for Sensor Scheduling

In this paper, we propose two myopic sensor scheduling algorithms for target tracking scenarios in which there is a tradeoff between tracking performance and sensor-usage costs. Specifically, we consider the problem of activating the lowest cost combination of at most L sensors that maintains a desired squared-error accuracy in the targets position estimate. For sensors that provide position information only, we develop a binary (0-1) mixed integer programming formulation for the scheduling problem and solve it using a linear programming relaxation-based branch-and-bound technique. For sensors that provide both position and velocity information, we pose the scheduling problem as a binary convex programming problem and solve it using the outer approximation algorithm. We apply our scheduling procedures in a network of sensors where the sensor-usage costs correspond to network energy consumption. Our simulation results demonstrate that scheduling using binary programming allows us to obtain optimal solutions to scheduling involving up to 50-70 sensors typically in the order of seconds

Sensor Resource Allocation for Tracking Using Outer Approximation

In this letter, we consider the problem of activating the optimal combination of sensors to track a target moving through a network of sensors. Our objective is to minimize the predicted approximate error in the target position estimate subject to constraints on sensor usage and sensor-usage costs. We formulate the scheduling problem as a binary convex programming problem and solve it using the outer approximation (OA) algorithm. We apply the proposed OA scheduling method to the scenario of tracking an underwater target using a network of active sensors. We demonstrate using Monte Carlo simulations that our OA scheduling method obtains optimal solutions to scheduling problems of up to 70 sensors in the order of seconds

The use of particle filtering with the unscented transform to schedule sensors multiple steps ahead

In a multisensor network, sensor scheduling can be used to minimize the cost of resources and improve system performance. We propose a multisensor scheduling algorithm using a particle filter and the unscented transform for a target tracking application. Under the constraint that only one sensor may be used at each time step, we predict the expected cost multiple steps ahead. We achieve this using several sets of particles for each sequence of sensors and then choose the sequence that minimizes the predicted cost. An advantage of the proposed algorithm is that it can incorporate arbitrary cost functions. Monte Carlo simulations, using squared error as the cost function, demonstrate the improved target tracking performance achieved with sensor scheduling.

Efficient search strategies for non-myopic sensor scheduling in target tracking

We propose two tree pruning algorithms to reduce the computational complexity of non-myopic sensor scheduling for target tracking. We consider a mobile bearings-only sensor that chooses from a finite set of possible moves at each time epoch. The scheduling objective is to select the sequence of sensor moves to minimize a tracking cost over M > 1 future time epochs. Tracking is performed with a particle filter, and expected future costs are calculated using an unscented transform with the particle filter. Simulation shows that the two algorithms significantly reduce the time and memory requirements compared to exhaustive search.

Sensor Scheduling using a 0-1 Mixed Integer Programming Framework

In this paper, we propose a novel myopic sensor scheduling methodology for tracking a target moving through a network of energy-constrained acoustic sensors. Specifically, we address the problem of activating the minimum-energy combination of sensors in a network that maintains a desired squared-error accuracy in the targets position estimate. We first formulate the scheduling problem as a binary (0-1) nonlinear programming (NLP) problem. Using a linearization technique, we then convert the 0-1 NLP problem into a 0-1 mixed integer programming (MIP) problem. We solve the reformulated 0-1 MIP problem using a linear programming relaxation based branch-and-bound technique. We demonstrate through Monte Carlo simulations that our proposed MIP scheduling method is very computational efficient as we can find optimal solutions to scheduling problems involving 50-60 sensors with processing time in the order of seconds

Acoustic Echo Cancelation for High Noise Environments

Acoustic echo cancellation (AEC) is highly imperative for enhanced communication in noisy environments such as a car or a conference room. In this work, we present a dual-structured AEC architecture that improves both the convergence time and misadjustment of a conventional adaptive sub-band AEC algorithm in high noise environments. In this architecture, one part performs smooth adaptation while the other part performs fast adaptation; a convergence detector is implemented to facilitate switching between the fast and smooth adaptations. We propose the momentum normalized least mean square (MNLMS) algorithm for smooth adaptation and we implement the NLMS algorithm for fast adaptation. The current architecture provides up to 3-4 dB echo reduction improvement over a conventional adaptive subband AEC algorithm and it helps minimize near-end distortion and artifacts in the post-processed AEC output