Options
2003
Conference Paper
Title
On a smart structure variable supervisory control concept for humanoid robots
Abstract
The increasing complexity of robot tasks both in the industrial manufacturing area and more and more in the private human environment requires a new generation of safe and flexible robots. More intelligence can be achieved by distributed multi-sensorics for the communication and interaction with its environment a close partnership and physical interaction between man and robot for solving complex tasks as well as learnability of the robot control system. Enhanced safety and flexibility of the robot requires the introduction of advanced structure variable supervisory control concepts which permit a dynamical situation responsive reconfiguration of case specific hardware and software components (e. g. sensors, sub-controllers). Situations and events which require a dynamic reconfiguration of the control algorithm may be e.g. special process phases and fault events. Robot control systems available on the market cannot comply with such advanced requirements. In order to cope with such new challenges a hierarchically structured Neuro-Fuzzy (NF) based supervisory control concept is proposed which has been successfully applied by IITB for the flexible automation of a complex industrial batch process. It relies on the decomposition of the complex control problem into a sequence of smaller, more transparent phase specific subcontrol problems. In the upper control level the actual process phase is identified by NF based multi-sensor diagnosis followed by a heuristic Fuzzy based decision and activation of the corresponding phase specific sub-controller. The lower level of the control structure consists of different low level sub-controllers which are optimized with respect to specific process phases. Principally the various low level sub-controllers may have any structure (Fuzzy control, model-based or hybrid). The aim of this paper is to present a new structure variable supervisory control concept for smart multi-sensory robots as well as to discuss about first experimental results from an ongoing project at the robot laboratory of the IITB.