Nao robot path planning method based on deep Double-Q network

Abstract

The invention relates to a Nao robot path planning method based on a deep Double-Q network, which aims at an uncertain environment, utilizes local information acquired by a robot to realize active obstacle avoidance in an indoor environment, plans a path to reach a set target point, and improves the generalization ability of a common path planning algorithm. Through preprocessing training environment data to approach local environment information which can be acquired by a robot in a real scene and setting a proper state space, a proper action space and a proper reward function, the method can directly utilize the acquired local environment information to perform effective obstacle avoidance and path planning in an unknown environment; the method overcomes the defect that a traditional algorithm needs to obtain sufficient environmental information and perform modeling again to explore the environment to a certain extent, improves the generalization ability of the algorithm, and is beneficial to solving the problem of path planning in an unknown environment.

Description

technical field [0001] The invention belongs to the field of deep reinforcement learning and path planning, and relates to a Nao robot path planning method based on a deep Double-Q network. Background technique [0002] Path planning refers to the technology of finding a collision-free path from the initial state (including position and attitude) to the target state (including position and attitude) in an environment with obstacles according to certain evaluation criteria. [0003] At present, path planning algorithms can be roughly divided into: classical algorithms and artificial intelligence algorithms. Traditional path planning algorithms mainly include: simulated annealing algorithm, artificial potential field algorithm, tabu search algorithm, etc. With the rise of artificial intelligence, because of its certain self-learning, self-renewal and memory capabilities, many path planning algorithms based on artificial intelligence have been proposed, typical ones are: ant c...

AI Technical Summary

Problems solved by technology

Most of the above algorithms need to have a full understanding of the surrounding environment information of the robot, and then carry out navigation planning based on certain rules. The actual application environment changes a lot and does not have the conditions to fully obtain environmental information, so it is necessary for the robot to be able to identify routes from unknown environments, be able to cope with different work scenarios, and complete path planning tasks

Images