Adaptive Dynamic Programming Approaches For Collaborative Target Tracking In Energy Harvesting Wireless Sensor Networks

Collaborative target tracking is one of the most important applications of wireless sensor networks(WSNs),in which the network usually relies on sensor scheduling to balance the tracking accuracy,energy consumption and network lifetime,due to the limited network resources for sensing,communication,and computation.With the development of energy acquisition technologies,the energy harvesting based WSNs have aroused people's attention,which can overcome the limitation of battery energy for the nodes in WSNs,where theoretically the lifetime of the network could be extended to the infinite.However,energy-harvesting WSNs pose new technical challenges for collaborative target tracking on how to schedule sensors over the infinite horizon under the restriction on limited sensor energy harvesting capabilities.On the other hand,as a new optimization algorithm,adaptive dynamic programming(ADP),which combines with some techniques such as neural networks(NNs),adaptive critic designs and reinforcement learning,provides an effective approach for the optimal control problem over the infinite horizon.This dissertation proposes the following optimal sensor scheduling approaches based on the global performance optimization for the collaborative target tracking in the energy harvesting WSNs over an infinite horizon:(1)For the single-target tracking problem,an ADP-based multi-sensor scheduling approach is proposed.Extended Kalman filter is adopted for the target state prediction and estimation.The performance index consists of the energy consumption and the tracking performance.The sensor scheduling is approximatively optimized with the performance estimated by a NN.Meanwhile the optimality of the proposed algorithm is analyzed.(2)For the single-target tracking problem,a multi-step prediction based ADP approach is proposed to schedule multiple sensors,which obtains the previous optimal multi-step sensor schedule by searching a decision tree and updates the remaining infinite-step performance by a NN.(3)For the multi-target tracking problem,an ADP based approach is proposed to schedule the targets and sensors.The performance index consists of the energy consumption and the weighted sum of the targets' tracking performance.The approximate control policy is iteratively obtained through the performance estimated by a NN.For each time step,one target is chosen as the observed object,for which multiple sensors are determined to perform the observation.