Abstract: In order to solve the problem of deep Q-network, The path planning task of DQN faces problems such as low sample efficiency, slow convergence speed, and large memory requirements, making it faster and better to accurately track targets in complex environments. On the basis of deep reinforcement learning technology, the D3QN algorithm is proposed to achieve path planning for targets on mobile sensors. The D3QN algorithm has three advantages. Firstly, it improves sample efficiency through a priority experience replay mechanism; Secondly, by using different value functions for the selection and evaluation of actions, the impact of overestimation can be reduced; The third is to improve convergence speed and effectiveness by changing the network architecture. And simulation verification was conducted using the GNU Radio software platform and Python language. Experiments have shown that compared to the DQN method, The method of D3QN can effectively reduce system energy consumption, improve real-time processing of tasks, and provide significant speed improvement and robustness.