Research Initiation Award: Task-Level Programming for Manipulation of Multiple Robots

任务级机器人编程包括语言处理、运动学、动力学、路径规划、控制、避碰和多机器人感知融合的计算，采用工业标准的vmebus分布式计算系统，由多个标量和矢量处理器组成，通过粗粒度矢量化和并行化，实时实现多机器人操作器的任务级机器人编程。该计算机系统由现成的商用VME板组装而成。因此，该项目的成果可以很容易地转移到实际工业应用中，从而对任务级多机器人操作系统的发展作出有意义和重大的贡献。由于编程环境是C/Unix兼容的，机器人编程可以利用现有的大量熟练的C程序员和Unix用户。本研究研究了多机器人操作的任务级编程，并研究了其在力和运动协调、碰撞检测和避免碰撞以及传感器融合方面的基本原理和算法。研究发展了基于传感器的多机器人操作与力协调理论，实现了不同机器人之间广泛的信息共享；发展了运动物体碰撞检测与避碰算法；发展了力与视觉传感器与机器人运动的集成技术。开发了一套标准的任务级函数。通过这些任务级功能，可以方便地完成复杂的机器人任务。通过计算机模拟和物理实验验证了理论、算法和任务级机器人编程。

The task-level robot programming with computations of language processing, kinematics, dynamics, path planning, control, collision avoidance, and sensory fusion for multiple robot manipulators will be achieved in real time via coarse-grain vectorization and parallelization using an industrial standard VMEbus-based distributed computing system consisting of multiple scalar and vector processors. This computer system is assembled with off- the-shelf commercial VME boards. The results of the project can, therefore, be easily transferred for practical industrial applications and should thus make a meaningful and significant contribution towards the development of task-level multi-robot manipulation systems. Since the programming environment is C/Unix compatible, the robot programming can draw on the large existing body of skilled C programmers and Unix users. This research investigates task-level programming for manipulation of multiple robots, and studies its underlying principles and algorithms for force and motion coordinations, collision detection and collision avoidance, and sensor fusion. The research developes a theory for sensor-based multi-robot manipulation and force coordination with extensive information sharing among different robots, algorithms for collision detection and collision avoidance of moving objects, and techniques for force and vision sensor integration with robot motions. A set of standard task-level functions are developed. Through these task- level functions, a complicated robot task can be conveniently accomplished. The theory, algorithms, and task-level robot programming are validated through computer simulations as well as physical experimentations.