Robot path planning method and planning system

Abstract

The invention discloses a robot path planning method. The method comprises the following steps: 1, constructing a two-dimensional environment grid map; 2, forming an evolutionary strategy neural network, wherein the input of the evolutionary strategy neural network is the robot position, and the output of the evolutionary strategy neural network is the probability that the robot executes each action at the position; 3, setting iteration parameters and initializing the iteration parameters; 4, generating a strategy population; 5, calculating attributes of the strategy population, and establishing an S set and a U set; 6, constructing a deep reinforcement learning network; 7, updating parameters of the deep reinforcement learning network by using U, and updating U and S; 8, adding the Gaussian noise to parameters corresponding to the strategy in the step S to generate a new non-elitist strategy, wherein the new non-elitist strategy and a previous-generation elitist strategy form a new strategy population; 9, when the number of iterations reaches the maximum value, by using the optimal strategy in the strategy population as a result, acquiring the action of each step of the robot from the starting position to the destination position, or otherwise, returning to the step 4 to carry out next iteration. According to the method, empirical data are repeatedly used for strategy improvement so as to quickly find an optimal path.

Description

technical field [0001] The invention belongs to the technical field of robot navigation, and in particular relates to a method and system for path planning of a mobile robot. Background technique [0002] In recent years, in order to cope with the growing application requirements and the improvement of the adaptability and flexibility of mobile robots in unfamiliar environments, scientists have conducted research on the architecture and planning of local trajectories for mobile robots. Although many effective methods have been proposed for mobile robot navigation in known environments, these methods are often difficult to apply to autonomous navigation in unknown environments. Deep reinforcement learning algorithms have been successfully applied to a range of challenging control tasks, but their lack of effective exploration limits the applicability of adopting deep reinforcement learning in path planning. The main problem of existing robot navigation methods is that in an ...

AI Technical Summary

Problems solved by technology

However, evolutionary strategies often suffer from high sample complexity and are difficult to solve problems that require optimizing a large number of parameters

Images