A motion planning method for underwater robot based on multi-constraint objectives

Abstract

The invention discloses a motion planning method for an underwater robot based on a multi-constraint objective, which belongs to the field of machine learning and motion planning of an underwater robot. Model construction stage: convert the robot obstacle avoidance sonar signal and the flow velocity signal of the flow velocity sensor into the current environment; establish a discrete action space according to dynamic constraints; establish a reward function with underwater obstacles as constraints; Cove's decision-making process, which establishes the foundation for algorithm implementation; training phase: training based on the Q-learning algorithm, in the current environment, based on the greedy strategy to execute actions, each step of strategy execution, based on the original strategy to evaluate and update the strategy, improve the strategy until it adapts environment to achieve planning goals. The invention considers multi-constraint targets such as water flow, obstructions, and targets, and combines the reinforcement learning method with underwater multi-constraint targets to realize the motion planning of underwater robots, which has strong real-time performance and can be applied to various surroundings.

Description

technical field [0001] The invention belongs to the field of machine learning and motion planning of underwater robots, and in particular relates to a motion planning method for underwater robots based on multi-constraint targets. Background technique [0002] Intelligent underwater robots have broad application prospects in marine scientific research, marine development, underwater engineering, and military affairs. Intelligent underwater robots generally work in complex marine environments. In order to better complete various missions and ensure their own safety, they need to have autonomous motion planning capabilities in unknown environments, and be able to avoid obstacles, Navigate to the target point. [0003] Traditional underwater robot motion planning technology needs to build a global map in advance. When the environment changes, the connection model needs to be re-established, which has poor adaptability and poor practicability. Reinforcement learning is an uns...

AI Technical Summary

Problems solved by technology

[0004] Reinforcement learning has been proven to be used in underwater robots, but the traditional method of motion planning for underwater robots based on reinforcement learning considers a single constraint goal, and does not take into account multiple constraints such as water flow constraints, target constraints, and obstruction constraints. The Influence on the Motion of Underwater Robot under the Condition of Target Constraint

Images