| Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 99 === A prototype of indoor service robot system using sparse image waypoints to navigate is developed. Upon deployment, the robot visually learns the work environment by traversing along the topology of the environment and automatically computing missing waypoints between service locations. Five types of reference images are used by the robot. While approaching a service target, the robot adopts a fast motion mode when traveling between intermediate waypoints, then switch to accurate motion mode when it is close to the target. Also an alternative navigation approach is proposed which pilots the robot back to the home position by an modified log-polar matching method for a discernible feature environment. The approach reserves processing time for the pre-training phase and reduces the size of the reference image database. By applying the fast mode to pass intermediate waypoints, the navigation time is shortened about 19 percent compared with the previous work of Fu et al. With the reference images generate by the proposed method and the fast navigation mode, a 160 meter navigation route can be finished in about 11 minutes with tolerable errors.
|
|---|
