Physics-based high-performance bionic underwater robot simulation training platform construction method

Abstract

The invention provides a physics-based high-performance bionic underwater robot simulation training platform construction method which can be used for training a fish-shaped underwater robot. The platform is composed of a robotic fish modeling module using a joint body with skin, a GPU-based high-performance local bidirectional fluid-solid coupling simulation module (capable of processing a finite field and an infinite field), and a reinforcement learning module. A plurality of bionic robot models are provided, a series of underwater reference control tasks are formulated, and an existing reinforcement learning training method is applied to obtain a corresponding control strategy. According to the method, high-performance fluid simulation requirements of a finite domain and an infinite domain can be processed, a model of any bionic underwater robot can be built, an underwater control task can be customized, and intelligent strategy training, migration and integration can be realized.

Description

technical field [0001] The invention belongs to the fields of computer graphics, fluid simulation, bionic robots and machine learning, and relates to a construction method of a physics-based high-performance bionic underwater robot simulation training platform. Background technique [0002] Bionic underwater robots have superior maneuverability, propulsion efficiency and realistic visual appearance. These advantages have prompted the research, production and application of bionic soft robots and bionic fish-like robots in the past few years. Underwater robots have important applications in many aspects, such as marine education, navigation and rescue, seabed exploration, scientific surveying, etc. However, due to the high cost of hardware, unstable environment, endurance, slow data collection and many other practical problems in studying the control mode and motion mode of robots in a real environment, training robots to obtain motion control strategies that approximate or ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is: to propose a high-performance bionic underwater robot simulation training platform based on physics, to solve the problem of imperfect modeling of underwater robots in the prior art, lack of fluid simulation training environment, limited fluid simulation area, low performance, and intelligent Difficulties in task generalization, knowledge transfer, and multi-skill combination faced in the context of learning

Images