机器人-英语原文.docVIP

WHISKER BASED MOBILE ROBOT NAVIGATION David Jung and Alexander Zelinsky Intelligent Robotics Laboratory Computer Science Department, University of Wollongong Wollongong, NSW 2500, Australia. djung@cs.uow.edu.au 1. INTRODUCTION One of the requirements of our research, and a common requirement for an autonomous mobile robots, is the ability to navigate competently in an unstructured environment. This problem has received much research attention in the robotics community. A major difficulty is that there are two competing aspects to navigation. A robot must be able to react very quickly to its environment to prevent collisions with static and dynamic obstacles. Also, in order to achieve useful purposive navigation a robot must use a map-like representation of its environment. This implies some sort of limited cognition (as defined by [McFarland 91]). The former implies minimal processing to meet the real-time requirements, while the later requires more processing than can typically be achieved for real-time response. The basic control design approaches can be broadly divided into four types. These are defined by[Mataric 92a] and we briefly reiterate them and their shortcomings here. The purely reactive approaches use a mapping from sensor sets to associated actions; a set of rules [Brooks 87]. The planner-based strategies originated with the symbolic AI community and employ a sense-plan-act cycle. The plan stage uses cognitive techniques to reason about a symbolic world model. There also exist hybrid systems which employ reactive components beneath planner-based systems to provide the benefits of both. Another approach is behaviour-based, which uses a set of interacting distributed concurrent behaviours, each of which may incorporate memory and learning of environment representations. Figure 1 - Yamabico Robot with Whiskers The purely reactive approach achieves robust realtime performance, but the tasks that can be achieved are limited because of the lack of any cogni