Design and control framework for cooperative mobile robot collectives
Tang, Chin Pei
MetadataShow full item record
This dissertation examines development of a framework to achieve cooperative payload transport and manipulation by teams of multiple wheeled mobile robots. Interest in such cooperative systems typically arises when certain tasks are either too complex to be performed by a single agent, or when there are distinct benefits in terms of redundancy, robustness and reliability that accrue by many simple cooperative robotic modules. However, the mechanical capability of both the individual modules as well as their interactions can significantly affect the overall system performance. Hence, we address various fundamental issues in terms of the design analysis, motion planning and control to equip the robotic modules with mechanical level system-based thinking to realize the physically tight cooperative payload transport task. We begin by the preliminary investigation of the development and evaluation of alternate distributed formulations for dynamic simulation of constrained multibody systems. Two computation methods are compared and evaluated in detailed using the benchmark problem of a four-bar linkage. Although the examination is preliminary, the contribution of this work is to establish a framework to simulate larger cooperative systems with more constraints in a distributed manner. We then investigate the optimal relative layout for members of a team of differentially-driven wheeled mobile robots (DD-WMRs) moving in formation for ultimate deployment in cooperative payload transport tasks. The contribution of this work is on the modeling of such formation and the development of the optimal motion plans in an "energy efficient" manner globally over a prescribed trajectory using the tools from differential geometry. This work provides the global optimal motion plans that improves the formal local optimization framework presented in , which results in smoother collective maneuvers. However, the various individual modules contribute both constraints and capabilities to the cooperative system. Thus, we also seek to examine the performance of such cooperative system composed from the capabilities of the individual wheeled mobile manipulator agents using the tools from screw theory. We argue that a proper design can provide the cooperative system capabilities to: (a) accommodate/create arbitrary payload motions, and (b) support/resist arbitrary payload forces. The contribution of this work is to carefully examine the system level performance in terms of reconfigurability and actuation requirements. Finally, we examine the use of differential flatness to provide a convenient and formal method to integrate both (point-to-point) motion planning and control of nonholonomic mobile robot systems. The contribution of this work is to develop such control framework for wheeled mobile manipulators and evaluated in the hardware-in-the-loop experimental framework.