Virtual Assembly Path Based On Deep Reinforcement Learning Research On Planning Method

Virtual assembly is an important research direction of virtual manufacturing.Virtual assembly can verify the rationality and assemblability of electromechanical products design,reduce the production cost and development cycle of products,and thus improve the competitiveness of products.Path planning is an important development direction of virtual assembly.The study of automatic path planning technology in virtual assembly environment is of great significance for the assembly path design of complex assemblies.The assembly path planning starts from the assembly starting point of the product,and solves the path according to the characteristics of the assembly environment and the relative positional relationship between the components to be assembled and other components,and finally obtains a collision-free path that satisfies the assembly requirements.The virtual assembly scene is built on the computer to realize the planning of the assembly path,and the assembly effectiveness of the assembled product parts can be verified in time.However,path planning is mainly used in the field of robots.The simple environment is identified by sensors,and the path is planned according to environmental modeling.However,virtual assembly has a large number of complex environments,which are not easy to be modeled and contain narrow space.Traditional path planning algorithms cannot guarantee the complexity and planning efficiency of complex environments,so it is not suitable for direct use in the complex environment of virtual assembly.Aiming at the complex environmental accessibility caused by the relatively narrow free space in virtual assembly,the deep reinforcement learning of optimal action sequence decision learning is proposed by means of tracking and feedback.This paper implements the path planning of virtual assembly based on deep reinforcement learning,and mainly completes the following aspects:(1)Introduced the development and research status of the path planning field.By analyzing the research results and practical problems at home and abroad,the problems solved in this paper are summarized: the complex environmental access problems caused by the relatively narrow free space in virtual assembly.(2)Introduce the basic theory of deep learning and reinforcement learning in detail.On the basis of analyzing the algorithm principle and applicability of deep learning and reinforcement learning,this paper proposes to use deep reinforcement learning to solve the path planning problem of virtual assembly.The combination ofdeep learning perception and intensive learning decision-making ability is suitable for solving sequence decision problems.(3)Converting the path planning problem of virtual assembly into a sequence decision problem of path finding.It is theoretically verified that the deep reinforcement learning proposed in this paper can train the deep network to solve the automatic path finding problem.(4)In order to verify the experimental effect of fuzzy Bayesian-depth Q network algorithm in virtual assembly environment,the complex environment of virtual assembly is simulated.The algorithm is studied in the simulation environment to ensure the feasibility of the method.Effectiveness.It has better passability and planning efficiency in a complex environment with a narrow space for virtual assembly environments.