Two important research issues in mobile robot research are navigation and recognition. The goal of this work is to provide a robot with the abilities to navigate in an iudoor enviromnent and to identify openings like a door as au erltrauce or an exit. Our research platform is the Labmate hardware purchased from Transitions Research Corporation, Connecticut, USA. We have built on top of the Labnate a souar proximity system aud a vision system. No global map or world model or higher-level knowledge is stored in the system. Vision is used to guide the robot to move towards a door. In the visiou module, vertical lines are the primary features used for door detection. It also gives au approximate direction of the door detected so as to direct the moveineut of the robot. Primitive obstacle av...[ Read more ]

Two important research issues in mobile robot research are navigation and recognition. The goal of this work is to provide a robot with the abilities to navigate in an iudoor enviromnent and to identify openings like a door as au erltrauce or an exit. Our research platform is the Labmate hardware purchased from Transitions Research Corporation, Connecticut, USA. We have built on top of the Labnate a souar proximity system aud a vision system. No global map or world model or higher-level knowledge is stored in the system. Vision is used to guide the robot to move towards a door. In the visiou module, vertical lines are the primary features used for door detection. It also gives au approximate direction of the door detected so as to direct the moveineut of the robot. Primitive obstacle avoidance algorithms, which only enable the robot to move away from obstacles are not sufficieut for our purpose because these algorithms do not take the goal into consideration. Some sophisticated algorithms like the poteutial field method could take the target point into consideration, but the target has to be specified in a coordinate system with respect to the environnlent. Instead of Cartesian coordinates, a robot center polar coordinate system is used in specifying the direction of motion. A navigation algorithm which guides the robot back to the original path is proposed. A cooperative algorithm using both sonar and vision information to achieve the door entrance task is also presented. These algorithms have been tested on our robot and three examples are discussed.