High Reliability Of Autonomous Mobile Robot Software Architecture Research

Although increasing applications of autonomous mobile robots (AMR) come forth in this century, reliability technology in this field is not yet advanced enough to provide those applications with a robust robot system, since sensor-processor-actuator systems such as AMR present a rich, complex problem domain that can exhibit significant levels of noise and uncertainty.With the goal of bringing some objective grounding to this important area of research, this thesis presents a case study of reliability problems in AMR. A multi-agent based AMR system, implemented on physical vehicle, is used as testbed. Analysis of test data collected from ground test identifies that the unreliable results from two main reasons: large latency to response urgent perceptional event and exception of inner modules (agents).To improve the reliability of AMR, two methods have been developed:(1) By casting reliability problem in the well-understood framework of software fault-tolerance, a vehicle health system has been developed to monitor the inner modules and diagnose based on rough set technology for rule reduction;(2) By implementing hybrid architecture, a latency-shorten method has been developed to optimize the performance of MAS-based autonomous mobile robot under urgency.The experimental result shows that the two methods improve the reliability effectively.