Proceedings PaperBlank-map orienteering for a mobile robot using certainty grids
|Format||Member Price||Non-Member Price|
Certainty grids have been shown to be an effective method of generating accurate map data from incomplete sensor data. This is put to the test as we use a robot based certainty grid to maintain map information generated from eight fixed sonars to compare three robot navigators. Due to the low resolution and single sensor type, the certainty grid includes a variety of averaging and weighting techniques to improve sonar accuracy and reduce noise. The navigators are constrained by two design parameters: they should not use domain specific knowledge and the navigators and mapper are independent. Navigation decisions are based solely on the internal map. Each navigator uses a weighting function to determine a potential for each grid element and navigates by minimizing the potential over the robot's immediate surroundings. Local route selection is performed in real time while traveling as the local navigator continuously re-evaluates the path with new information from the certainty grid. The navigators differ in their methods of global route selection. One uses intermediate destinations and backtracking to handle dead ends. The other two incorporate dead end information directly into local route selection, one with intermediate destinations and the other without them.