Research On Modularized Collaborative Position/Force Control For Environmental Constrained Reconfigurable Robot

With the development of science and technology, robot is playing an increasingly important role in the information age. Traditional robot is often designed to meet the need of a specific task. Although it is possible to realize the various process operations through changing its own control program, the fixed configuration still greatly limits the application range. Therefore, modular and reconfigurable robot(MRR) with strong ability to switch configurations and adapt to different environment and mission requirements emerges. Accordingly, the reconfiguration of MRR contains double reconstruction on both mechanical structure and control system. In view of its high flexibility, strong environmental adaptability and low cost, etc., MRR has potential applications in assembly industries, medical, military, disaster relief, deep space exploration or even nuclear accident treatment. So the research on MRR is meaningful for both theoretical analysis and practical applications.Since the emergence of MRR, the tracking control problem in free space has caused wide attention of scholars and the research results are relatively mature. When MRR is accomplishing tasks such as precision equipment assembly, metal surface grinding and polishing and components of the welding, the endpoint will contact to the environment inevitably. In such cases, it need to analysis the position/force control problem of the constrained MRR. Referring to all of the available literature concerned, the study on position/force control problem of the constrained MRR is still rare at home and abroad. In addition, MRR may need transform its configuration to adapt to different mission requirements during the mission. So for the essential modular design ideas of MRR, the decentralized control strategy is more suitable for its control problem. However, the model decomposition, namely how to use only local joint information to map the endpoint contact force to each joint, is always a bottleneck which limits the application of decentralized control strategy to the position/force control problem of the constrained MRR. In this article, on the basis of existing research findings, the position/force control problem of the constrained MRR has been analyzed. The research mainly covers the automatic dynamic modeling approach of the constrained MRR, adaptive neural network position/force control of the constrained MRR based on dynamic model decomposition, as well as decentralized adaptive neural network sliding mode position/force control of the constrained MRR based on local information.The organization of this thesis is as follows, in which the main works are included.1. Research background and significance of the study are elaborated first, then summarize and analyze the research status and contents of MRR.2. Based on Newton-Euler iterative algorithm, the dynamic model of MRR in free space is established. The relationships between the contact force and force/torque in joint space are derived based on the virtual work principle. The dynamic model of the constrained MRR is finally formulated.3. The position/force control problem of the constrained MRR is analyzed. Based on a nonlinear transformation, the dynamic model of MRR is divided into the position and the force subsystem, respectively. Then neural network controllers are introduced to approximate the uncertainties and nonlinear term in each subsystem and the corresponding estimated errors are compensated adaptively. The Lyapunov stability theory guarantees the stability of the closed-loop constrained MRR system.4. Combined with the more suitable algorithm of its design ideas, the decentralized position/force control scheme of the constrained MRR based on local information is discussed. The contact force is mapped to joint space via Jacobian matrix which includes the i-th joint’s real-time information and other joints’ desired information. Based on model decomposition technique, the dynamics can be represented as a set of interconnected subsystems. By adjusting the converted torque of each subsystem in joint space, the endpoint contact force in work space is controlled indirectly. The subsystem dynamics is represented by the neural network controller and the generality of the given control parameters under different configurations is ensured. And the sliding mode controller is designed to compensate the interconnection and estimate errors. In simulation part, two MRRs with different configurations are employed to show the effectiveness of the proposed decentralized position/force control scheme.Finally, the conclusions and prospect of future research are given at the end of this paper.