Robot motion planning method and system based on graph Wasserstein self-encoding network

Abstract

The invention discloses a robot motion planning method and system based on a graph Wasserstein self-encoding network. The method comprises the following steps of S1, constructing the graph Wasserstein self-encoding network (GraphWAE); S2, conducting non-uniform sampling distribution representation learning based on the GraphWAE; and S3, conducting robot motion planning based on the GraphWAE. Compared with the prior art, the robot motion planning method and system have the beneficial effects that the non-obstacle area in the robot configuration space is represented and learned through the GraphWAE, and serves as a sample generator of a mainstream sampling planning algorithm, the planning exploration process is guided to be carried out in the non-obstacle area, the planning time is shortened, and the planning path quality and the success rate are improved.

Description

technical field [0001] The invention relates to the field of intelligent robots, in particular to a robot motion planning method and system based on graph Wasserstein self-encoding network. Background technique [0002] With the rapid development of intelligent robots, especially service robots (including daily service robots, special service robots, etc.), the scenes and environments faced by robots have become more and more complex. When performing various complex tasks, efficient motion planning method is very important. [0003] Common problems in the current industry when robots are performing motion planning include: [0004] 1. Motion planning takes a long time. Due to the current mainstream planning methods such as RRT*, PRM*, FMT, etc., for complex obstacle scenarios, the planning time is often long, which is difficult to meet the needs of robots to perform actual operation tasks. [0005] 2. The planning path is poor. The quality of the planned path mainly depe...

AI Technical Summary

Problems solved by technology

Due to the current mainstream planning methods such as RRT*, PRM*, FMT, etc., for complex obstacle scenarios, the planning time is often long, which is difficult to meet the needs of robots to perform actual operation tasks

[0005] 2. The planning path is poor

A poor planned path usually means that the planned path is longer and has a more complex representation in space

For mobile manipulators, a poorly planned path tends to consume more energy, and the planned path is too complex, which may cause some potential dangers

[0006] 3. Low planning success rate

[0008] This existing technology has the following disadvantages: 1) When the obstacle scene is relatively complex, CVAE cannot accurately characterize the non-obstacle area of ​​the configuration space, resulting in low non-uniform sampling efficiency, which cannot effectively shorten the motion planning time and improve the quality of the planned path and success rate

2) The training time of CVAE is long, and the training crash often occurs, which makes it impossible to carry out subsequent planning

Images