Legged robot motion control method and device based on meta-reinforcement learning and medium

Abstract

The invention discloses a legged robot motion control method based on meta-reinforcement learning. The legged robot motion control method comprises the steps of: constructing a robot motion simulation environment; generating a motion planning strategy and a motion control strategy according to a meta-reinforcement learning algorithm; acquiring a target motion track of the robot according to the motion planning strategy; and according to the motion control strategy, controlling the robot to move from an initial state to a target state along the target motion track in a simulation environment. According to the legged robot motion control method provided by the embodiment of the invention, the motion control strategy can be generated for the robot in a high-dimensional continuous state space, the robot can perform path planning in combination with external sensing information such as images and laser radars, and the strategy can adapt to a variable environment; and the applicability of a robot control strategy is improved.

Description

technical field [0001] The invention relates to the technical field of intelligent robots, in particular to a motion control method, device and medium of a legged robot based on meta-reinforcement learning. Background technique [0002] With multiple discrete footholds and redundant degrees of freedom, legged robots are more suitable for unstructured environments than wheeled or tracked robots. The movement of legged robots is generally divided into two parts: planning and control. For the planning part, graph search algorithms such as the traditional A* algorithm and breadth-first search algorithm can be used to find the shortest path in the entire state space, and the motion planning can also be modeled. Solve constrained optimization problems. For the control part, methods such as modern control theory, optimal control, and model predictive control can be used to control each mechanism to follow the desired trajectory. [0003] However, the traditional planning control ...

AI Technical Summary

Problems solved by technology

[0003] However, the traditional planning control method is difficult to solve in the continuous state action space, which makes it difficult for the robot to perform motion planning in combination with external sensor information such as images and lidar, and the control methods in the prior art are often only applicable to specific environments

Images