PARALLEL PROCESS DECOMPOSITION OF A DYNAMIC MANIPULATION TASK - ROBOTIC SEWING

Conventional robot control schemes are inadequate for dynamic tasks involving gross environmental uncertainty, and consequently, current robots are restricted to artificial structured environments. The task planner approach, as promoted by the AI community, is unsuited to tasks involving interaction with a dynamic environment. An alternative approach, which is based on concurrent control processes or behaviors, can enable a system to perform dynamic tasks in an uncertain environment with inherent error recovery and graceful degradation. Dynamic tasks, such as locomotion and mobile robot navigation, have decomposed into parallel processes, resulting in control systems that are simple yet robust enough to operate in realistic unstructured environments.This paper investigates parallel decomposition of manipulation tasks that involve interaction with a dynamic environment. A methodology for developing concurrent process controllers for such tasks is outlined. A concurrent process control system was developed for a robotic sewing cell that produced an edge seam on a fabric panel with arbitrary edge curvature. Although a conventional sewing machine sewed up the fabric, a robot arm manipulated the panel to modify its orientation and control fabric tension. The sewing task was decomposed into four concurrent processes within a superposition parallel architecture. Although each process was based on simple linearized models, the complete system demonstrated robustness to the unpredictable dynamic behavior of the fabric panel. © 1990 IEEE