A case study in distributed robot arm planning

This paper describes a case study in which an assembly planning problem is distributed across a computer network. A distributed robot arm planning system (RAPS) is used to define a solution to the blocks world problem. This problem is concerned with the formation of a plan to enable a robot arm to manipulate arrangements of equally sized cubes situated on a flat table. The simulation work is based on this problem. The paper presents results on two key issues that are relevant to designers of distributed planning systems, namely cooperation frameworks and the efficient exploitation of distributed hardware. Because the components of a distributed planning system rarely have sufficient information or ability to solve the problem as a whole uniquely, the issue of cooperation arises. R.G. Smith and D. Randall (1981) identify two methods of cooperation, the result sharing and the task sharing frameworks. The result-sharing framework describes the cooperation of independent planning components through the communication of partial results. Integration and convergence are achieved by each component's attempt to maintain an internal state consistent with its environment. The demonstration software that implements this framework is coded in the Occam concurrent programming language and is targeted at a network of transputers. RAPS is evaluated using simulation tools.<
>