Study On Dynamics Control For Reconfigurable Modular Robots Based On Local Information

Over the past decades, robots have been uniquely designed to perform a specific task in the limited environments. Although this design concept has resulted in robots which are well suited for the specific tasks, it could not provide the flexibility to adapt to a variety of tasks. In some condition, using robots with different parameters for a variety of task is possible when the task requirements are specified in advance. However, it is very difficult or impossible to design a single robot system to finish some unpredictable, nonstructural environment or variant tasks, such as nucleus scrap heap recycle bin, cosmic space station and lunar base etc. A reconfigurable modular robot consists of interchangeable links and joint modules with standardized connecting interfaces. It can be easily assembled and disassembled, so it can accomplish a larger number of classes of tasks through the reconfiguration of a small inventory of modules. Compared with conventional robots with fixed configuration, reconfigurable modular robots can be adapted to diverse task requirements and take advantage of low cost, easy maintenance, convenient modification, portability and durability against system malfunctions. With the rapid development of robot technology, reconfigurable modular robots have been applied to military, aviation and industry. Several prototypes of reconfigurable modular robots have been developed in some research institutions to study on relative theories. These theories include kinematics and dynamics automatic model generation, software and control structure, kinematics calibration, optimal assembly configuration, and fault tolerance. However, the formalization of a generalized control scheme for such a reconfigurable modular robot is more difficult than conventional robots due to its flexibility in configuration. This thesis focuses on the kinematics and dynamics automatic model generations, inverse kinematics solution, distributed control, decentralized state feedback control and output feedback control, distributed fault diagnosis, decentralized fault tolerant control. Main content and innovation are as follows.A set of conceptual robot modules is introduced. The reconfigurable modular robot assembly configuration is represented by a configuration representation matrix, the kinematics and dynamics are generated based on the configuration representation matrix. Because inverse kinematics of reconfigurable modular robot has not an exclusive solution, the simulated annealing based genetic algorithm is introduced in optimizing the pure position problem, pure orientation problem and position orientation problem of inverse kinematics.A distributed adaptive sliding mode control based on modular information flow is proposed for reconfigurable modular robot. The controller of reconfigurable modular robot is considered as a distributed computing network, in which joint modules are considered as corresponding computing nodes. These nodes have functions of communication, sensing and controller. Each node uses the information of the corresponding joint and the adjacent joint in network to generate dynamics of the subsystem through recursive equations based on geometric formulation for the dynamics of rigid body, and then a controller is designed for each subsystems, all of which will constitute a modular control network to achieve the stable and reliable trajectory tracking control of a reconfigurable modular robot. Note that the computing error in the subsystem model is inevitable due to the communication delay and uncertain parameters. In order to handle the modeling error, an adaptive sliding mode controller is designed to estimate up-bound of the modeling error and compensate it adaptively.A distributed fault diagnosis method for reconfigurable modular robot is developed based on decomposing algorithm. Consider the localization of the fault diagnosis method, the distributed fault diagnosis outperforms the centralized fault diagnosis. By using the subsystem observer, the subsystem output can be estimated. The deviations between estimated and measured signals are utilized to train the neural network on-line which can identify the unknown fault occurring in subsystem, thereby, the distributed fault diagnosis can be achieved.A decentralized control scheme based on the adaptive fuzzy sliding mode control is proposed for the tracking problem of reconfigurable modular robot. In contrast to the previous decentralized controllers, the proposed control approach based on the nonlinear feedback control strategy which can adaptively compensate the subsystem nonlinearities. Due to the functional approximation capabilities of fuzzy systems, the dynamic models for each subsystem are not required to be linear parameterization of known nonlinear functions. For each subsystem, only local information are used to approximate subsystem dynamic model through a first-order Takagi-Sugeno fuzzy system, and then an adaptive sliding mode controller is designed to remove the effect of interconnection term and fuzzy approximation error. The proposed control scheme has several aspects. The first one is that decentralized control architecture can avoid difficulties in complexity of controller design, debugging, data gathering, and storage requirements. The second aspect is that, decentralized control architecture can instantly adapt to manipulator reconfigurations and can control reconfigured manipulator without having to adjust controller parameters. The last aspect is that, the controller of each joint module does not require any precise knowledge of the structure of the entire dynamic model.An observer based decentralized adaptive fuzzy controller for reconfigurable modular robot is proposed. The dynamics of reconfigurable modular robot is represented as a set of nonlinear interconnected subsystems. By separating terms depending only on local variables from the robot dynamics, each subsystem dynamic model can be formulated in joint space. The tracking problem is tackled with decentralized controller. The subsystem controller consists of adaptive fuzzy systems and robust term. By designing the state observer, the adaptive fuzzy systems which are used to model the unknown dynamics of subsystem and the interconnection term can be constructed using the state estimations. The effect of fuzzy approximation error is removed by employing the robust term. The sufficient conditions for stability of observer-controller, as well as the analytic relationship between the observer gain and tracking errors are given based on Lyapunov stability theorem.For reconfigurable modular robots, it is very challenging to design effective fault tolerant control due to diverse configurations and weak ability for communication. To satisfy the concept of modular software, a passive decentralized fault tolerant control scheme is proposed for tolerating actuator degradation at each joint module. Both indirect and direct adaptive fuzzy controllers are designed for each joint module. The fault tolerance is achieved at each joint module without requiring sub-dynamics or information about the other modules. All adaptive algorithms in the subsystem controller are derived form the sense of Lyapunov stability analysis, so that resulting closed-loop system is stable and the H∞tracking performance is guaranteed.The conclusion and the perspective of future research are given at the end of the paper.