Experiments in the integration and control of an intelligent manufacturing workcell

摘要

This thesis comprises the experimental development of an intelligent, dual-arm robotic workcell. The system combines a high-level graphical user interface, an on-line motion planner, real-time vision, and an on-line simulator to control a dual-arm robotic system from the task level. The graphical user interface provides for high-level user direction of the task to be done. The motion planner generates complete on-line plans to carry out these directives, specifying both single and dual-armed motion and manipulation. Combined with the robot control and real-time vision, the system is capable of performing object acquisition from a moving conveyor belt as well as reacting to environmental changes on-line. The thesis covers in detail four main topics: (1) System design and interfaces. The system is based on a novel interface-first design technique. This technique structures the complex command and data flow as combinations of three fundamental robotic interface components: the world-state interface, the robot-command interface, and the task-level-direction interface. (2) Network communication architecture. Complex distributed robotic systems require very complex data flow. A powerful new subscription-based, network-data-sharing system, was developed (and is being commercialized) that enables transparent connectivity. (3) Control system. The architecture and design of the hierarchical control system for the experimental dual-arm assembly workcell is described. (4) Path time-parameterization. A fast (linear-time, proximate-optimal) solution to the fundamental problem of time-parameterization of robot paths is presented. The design was verified experimentally in a dual-arm robotic workcell. Experimental results are presented showing the system performing complex, multi-step tasks autonomously, including dual-arm object acquisitions from a moving conveyor, object motion among obstacles, re-grasps, and hand-overs. All these tasks occur under task-level human supervision.