Mobile robot continuous control method based on non-map motion planner

Abstract

The invention brings forward a mobile robot continuous control method based on a non-map motion planner. Main contents of the method comprise the non-map motion planner, an asynchronous depth determination strategy gradient, reinforcement learning, an assessment network and a reward function. The process of the method is as follows: end-to-end training is performed by utilizing the non-map motion planner, and a transition function is found for the non-map motion planner in order to control frequency; an original depth is modified, and a strategy gradient is determined to be the asynchronous depth determination strategy gradient; the reinforcement learning is performed so that training and sample collection can be executed in parallel; and the motion planner is assessed by utilizing the assessment network, the reward function is defined, and whether a target is reached is checked. A high-precision laser range-finding sensor is used, a path can be accurately calculated, and efficiency is high; and any manual design and demonstration in advance are not needed, a feasible optimized path can be efficiently searched out, a robot is navigated to a target position, and the robot cannot collide with an obstacle in environment.

Description

technical field [0001] The invention relates to the field of robot control, in particular to a continuous control method of a mobile robot based on a mapless motion planner. Background technique [0002] With the development of science and technology, mobile robot navigation has gradually become one of the hot issues in the field of robotics and artificial intelligence. At the same time, it is also a manifestation of the intelligence level of fully autonomous robots. People hope that when working in an unknown environment, the mobile robot can obtain local environmental information based on its own sensors, build an environmental map autonomously, and plan a feasible path to reach the destination without collision according to the established map. In this way, mobile robots can be used in daily navigation, path planning and other fields, bringing convenience to people's travel and work. However, traditional methods use simultaneous localization and map creation to achieve n...

AI Technical Summary

Problems solved by technology

[0004] For problems such as time-consuming navigation, the object of the present invention is to provide a mobile robot continuous control method based on a mapless motion planner, using the mapless motion planner for end-to-end training, and finding a conversion function for the mapless motion planner to Control the frequency so that the robot can react immediately to new observations; modify the original depth-determined policy gradient to become an asynchronous depth-determined policy gradient; conduct reinforcement learning so that training and sample collection can be performed in parallel; use the evaluation network to control the motion planner Do the evaluation, define the reward function check to reach the goal

Images