2A1-H04 Negative Obstacle Detection and Drivability Judgement for Mobility User Support(Demonstration experiments of personal mobility robots in Mobility Robot Special Zone of Tsukuba-city)

Abstract

Methods of negative obstacle detection and risk detection are presented for supporting users of mobilities with a range image sensor. Risks on road surfaces are defined by subjective judgement from users that a surface is drivable or not so that risk detection can meet risk recognition of users. After the risks are modeled by offline learning by SVM of those judgements as well as range images, they are predicted online. Experiment results show that the risks are predicted correctly with 92 % of learning samples.