Agent-based management of a task-level multi-robot assembly cell.

Abstract

This dissertation presents a flexible coherent framework for managing the on-line, high-level operation of an assembly cell that has multiple robotic manipulators. This framework, which uses the concept of an agent as understood by the research subfield of multi-agent systems, supports the task-level scheduling and programming of assembly operations. The execution of an assembly operation was decomposed into the two distinct phases of part assembling and part fetching. Part assembling is concerned with how a robotic manipulator manoeuvres a part into its proper place within the overall assembly of a product. Part fetching has to do with suitably delivering a part to a robotic manipulator for part assembling. In this framework, the operation of a multi-robot assembly cell was distributed among four different types of agents: agent(s) responsible for part assembling, agent(s) responsible for part fetching, an agent to properly schedule the assembly operations, and an agent to control access to the shared physical space of the cell. Each of these agents was implemented in software using the object-oriented programming language C$\sp{++}$. The physical-level of this framework was evaluated by physically implementing a proof of concept, assembling agent. The system-level of this framework was evaluated by integrating the software agents into a discrete event simulation and then simulating a tractable and prototypical example for different operating circumstances.