Avoiding Negative Obstacles

Avoiding negative obstacles


Figure 18: VaMoRs with Autonav-Stereo-system from Sarnoff-Corporation and UBM demonstrating real-time negative obstacle avoidance at full video-rate in September 2000 on UBM-campus.

In a cooperation with US-partners in the framework of the joint project "AutoNav" (MoU) the Sarnoff Corporation, Princeton, NJ and UBM developed a vision system based on EMS-Vision and Sarnoff's pyramid chip technology, which for the first time allowed stereo interpretation at full video-rate for closed-loop avoidance of negative obstacles. VaMoRs drove at speeds up to 10 mph (16 km/h or 5 m/s) on grass surface of the former airport in Neubiberg, and stopped autonomously in front of the "Hachinger Bach" (a small river bed, upper images) and an artificial ditch of 60 cm width (lower images).

In the former case, VaMoRs was able to perform obstacle recognition under rather heavy pitching motion from uneven ground (crossing a dirt road), while in the latter case the parameters for obstacle detection had to be tuned such that relatively many false alarms had to be tolerated. Much higher computing power is needed and joint evaluation of color and texture features has to be done for achieving performance levels approaching the human one.