Unmanned aerial vehicle path planning method based on transfer learning strategy deep Q-network

Abstract

The invention discloses an unmanned aerial vehicle path planning method based on a transfer learning strategy deep Q-network (DQN). The method comprises the steps of: firstly carrying out the modelingand description of a dynamic environment where a UAV is located through adoption of a grid method, and building a state space model and an action space model of the UAV; secondly, initializing network parameters of a DQN and the current state of the unmanned aerial vehicle; training the DQN by adopting a return mechanism based on a social force model under the static environment model to obtain anetwork weight and an optimal action value; migrating the network weight and the optimal action value obtained by training in the static environment to the dynamic environment by using transfer learning, and continuing neural network training to obtain an action to be executed by the UAV; and finally, calculating the position of the unmanned aerial vehicle at the current moment to achieve path planning of the unmanned aerial vehicle in the dynamic environment. According to the method, the problems of the low DQN training convergence rate, the non-ideal path planning and the low success rate when the unmanned aerial vehicle performs path planning in a dynamic environment are effectively solved.

Description

technical field [0001] The invention belongs to the field of UAV path planning, and in particular relates to a UAV path planning method based on migration learning and DQN (Deep Q-Network), applying migration learning and deep reinforcement learning to carry out UAV path planning in a dynamic environment . [0002] technical background [0003] UAV path planning is a core issue in the field of UAV technology research, and related algorithms are developing rapidly. Traditional methods include: Dijkstra's shortest path search method (greedy algorithm), A* algorithm, ant colony optimization algorithm, reinforcement learning algorithm, etc. The core idea of ​​Dijkstra's algorithm is that the next vertex selected by each exploration is the point closest to the Euclidean distance from the starting point until the target is found. This method is only suitable for static maps with known overall information, and the efficiency is low; the A* algorithm is based on the Dijkstra method...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a UAV path planning method combining transfer learning and DQN algorithm, which can solve the problems of slow convergence speed and low success rate when DQN algorithm performs path planning in a dynamic environment

Images