TerrAdapt: A system for autonomous terrain classification and terrain-adaptive driving

The Massachusetts Institute of Technology and Quantum Signal LLC propose to research and develop the TerrAdapt system for terrain classification and vehicle driving control. This work will result in fundamental research advances, but will be practical enough for application on today"s COTS vehicle platforms. The system will maximize vehicle mobility due to the following key features: 1) TerrAdapt will merge both proprioceptive and exteroceptive sensor inputs to obtain a rich, physics-based prediction of terrain properties; 2) TerrAdapt will employ state-of-the-art methods for terrain classification to allow the system to learn from experience, and adapt to changing terrain conditions; 3) TerrAdapt will ensure safe vehicle control, and will enable automatic avoidance of both geometric and non-geometric hazards; 4) TerrAdapt will be applicable to both semi-autonomous systems and fully autonomous vehicles; 5) TerrAdapt will rely on low-cost COTS hardware, and be designed as an"add on kit"for integration with existing vehicle platforms.