Key Technologies For Reconfigurable Modular Robot System

Making machatronics products with intelligence, flexibility and individualization has currently become a main trend in products design. Reconfigurable modular robot system (RMRS) is made up of joint modules, link modules, end-effectors with different functions and size characteristics as well as the corresponding driving, controlling and communication modules. RMRS can be constructed various degree-of-freedom (DOF) and configurations robot system according to the user's demand and application situations. The reconfigurable modular design is aiming at a higher-level design based on the conventional modular design, which is not only manufactures-oriented but also users-oriented. Users can customize modules and freely design and construct their desired configuration. So it is significant to study the key technologies for RMRS, including its configuration design, kinematics, dynamics, path planning, trajectory planning and distributed control system.The reconfiguration and adaptability of RMRS are determined by the performance and matching ability of the connectors between the module systems. Based on a rational decomposition of manipulator structure, an experimental module system is designed and implemented, which comprises joint module, link module, gripper and quick coupling mechanism etc. The active joint modules are self-contained unit, which integrate driving, control and communication functions. A computer-aided configuration design is developed by means of an interactive environment based on intelligent design software DEST. Analytic hierarchy process (AHP) is used to evaluation and decision-making of configuration schemes.Because of the variety of the configurations of RMRS, to obtain the generalized inverse kinematics is a challenging task. Motion screw and product-of-exponentials (POE) formula are employed to model the kinematics of reconfigurable robots, and systematically simplification methods of POE formula are investigated, classifying and computing of subproblems are also implemented. Thus a generalized, decomposable and reusable approach for closed-form inverse kinematics of reconfigurable robots is developed. For the redundant robot and some non-redundant robot, their closed-form inverse kinematics are not existent, Jacobian matrix based numerical iterative method are investigated. In addition, automatically determining the workspace of reconfigurable robots is also important in customizing the configurations of RMRS, two algorithms, i.e. determination of boundary points by dimension degradation and binary search, determination of closed-curves in multi-region cross section of the workspace based on double linked list are proposed, which implemented the automatically calculation of the three dimensions workspace of RMRS.Based on twist and wrenches representations, the Lagrangian equations are used for RMRS dynamics anylysis. By making all robot link frames unified and applying POE formula, Jacobian matrix, wrenches and their transformation to the Lagrangian equations, an explicit closed-form of the Lagrangian equations is obtained, which facilitates computer implementation and configuration design, dynamics verification and control analysis as well as synthesis.A novel approach for collision-free path planning of a multiple degree-of-freedom articulated robot in a complex environment is proposed. Firstly, based on visual neighbor point (VNP), a numerical artificial potential field is constructed in Cartesian space, which provides the heuristic information, effective distance to the goal and the motion direction for the motion of the robot joints. Secondly, a genetic algorithm, combined with the heuristic rules, is used in joint space to determine a series of contiguous configurations piecewise from initial pose to its destination. Trajectory planning can be conducted either in the joint variable space or in the Cartesian space. Generally the path is constrained in Cartesian coordinates, while the actuator torques and forces at each joint is bounded in joint space. Hence, it becomes an optimization problem with mixed constraints (path and torque constraints) in the two different coordinate systems. An approach for robot plane curve trajectory planning is proposed, which satisfies the mixed constraints in both Cartesian space and joint-variable space. In the Cartesian space, the determination of control knot points, the time assignment among the knots, and the deviation estimation between the planned trajectory and the desired trajectory are discussed; and in the joint-variable space, using cubic spline polynomials to fit the segment between two adjacent knots and how to satisfy the joint physical constraints (the velocity, acceleration, and torque constraints) are solved. According to the deviation between the planned trajectory and the desired trajectory, knot points are inserted unevenly among the predetermined knots. Thus the deviation is decreased significantly by adding small number of knots.Distributed control system (DCS) is used in RMRS, because the joints are smart and the DOFs and configuration of RMRS are variable. The whole control system includes three hierarchies, the first hierarchy is path planning, which implements process planning and path assignments; the second one is trajectory planning, generating the sequence of time-based joint variables, and distributing the motion instructions to all joint controllers using Modbus protocol through RS-485 network; the third one is joint controlling, which receives the commands form PC and finishes joint motion control. The control circuit and driver circuit are designed, and the position control software is developed based on PID controller with anti-windup correction. Some comprehensive experiments demonstrate the system is effective and reliable.