Landmark Navigation for Unmanned Ground Vehicles

Mobile Intelligence Corporation (MIC) and Wayne State University propose a virtual teleoperation and visual landmark navigation system for UGV teleoperation in buildings, tunnels and caves. The challenge is search and navigation in complex structures with sparse visual cues, high-latency low-bandwidth communications, UGVs with low-cost vision sensors, and only intermittent static 2D visual feedback presented in non-real time. Our solution leverages prior research developments. The Virtual Teleoperation Interface enables the operator to define visual waypoints by driving a virtual UGV in a virtual world. The virtual world wraps a static image from the UGV onto a surface facet model. The UGV"s Scene Analysis function generates the surface facet model and identifies associated visual landmarks. The UGV uses Visual Landmark Referencing for self-location relative to visual waypoints. Our solution provides the operator with teleoperation control and visual feedback similar to traditional continuous teleoperation. It enables the UGV to follow the operators"navigation relative to visual cues and landmarks while communications are delayed or interrupted. It converges to standard teleoperation when communications bandwidth is available. The system will be integrated with MIC"s multi-platform OCU and UGV AutoNav products, for demonstration on a physical UGV. Integration will leverage MIC"s Virtual Systems Integration Laboratory software.