Intelligent Resource Collaboration in Mobile Target Tracking Oriented Mission-Critical Sensor Networks

Abstract

Mobile target tracking-oriented sensor networks are a special kind of Mission-critical Sensor Networks (MCSN), in which the various missions with the diverse priorities exist. However, it is challenging to achieve real time tracking while keeping the MCSN a long life time with limited energy provision in a complicated environment. In this paper, we develop a collaborative perception and intelligent scheduling scheme, which jointly optimizes the system responding latency and tracking accuracy with the constraint of the available energy. A new hierarchical architecture is proposed to realize the coupled function of perception and computation. In particular, the multi-node collaborative perception scheme is applied to obtain the excellent sensing capacity, and the Unmanned Aerial Vehicles (UAVs) play as the edge nodes to provide the computing service for those resource-constrained sensor nodes. To reach the sustained target tracking, we propose an intelligent tracking policy by exploiting the deep deterministic policy gradient (DDPG) method. Simulation results demonstrate that the proposed intelligent collaboration scheme can improve the tracking accuracy by 45.5% compared with the random selection scheme. The system cost is also reduced approximately by 17.3% while guaranteeing the tracking accuracy. INDEX TERMS Multi-target tracking, energy consumption, collaborative perception, deep reinforcement learning.

Intelligent Sensing and Computing in Wireless Sensor Networks for Multiple Target Tracking

“…Moreover, due to the extensive research on machine learning, lots of researches have upgraded and extended the algorithm, which can be widely used in WSNs [22]. In [23], they proposed a double time scale Q-learning algorithm with function approximation to alleviate the curse of dimension problems.…”

“…For our designed intelligent computing algorithm, we focus on the constructed hidden layers of neural networks and the number of iterations. Explicitly, the action a is sampled from a set of action A with a time complexity Oð1Þ for each iteration [22]. In the hidden layers, the main consideration focuses on back propagation.…”