Implementation of 3-D, LIDAR-Based Vision to Improve the Safety of Grading and Excavation Machines

Grading and excavation machines of the future will possess greater autonomy. For these machines to operate safely, especially in human-occupied environments, it will be necessary for them to possess real-time, automated perception. We are currently engaged in a cooperative project between NASA, Caterpillar Inc., and TRACLabs Inc., which includes the development, integration and testing of 3d vision on a Caterpillar-MTL (Multi-Terrain Loader). Recently, a milestone in this project was achieved when the MTL was intentionally driven (via remote control) toward a human subject. As the machine approached, it sensed the human and appropriately signaled with its headlights. If/when the human failed to evacuate the intended path, the MTL would then revert to "safe-mode" in which it appropriately halted. Herein we discuss the vision technology used in this experiment as well as adaptations to the MTL's control system which enabled modular integration. Critical features of the 3d, LIDAR-based, vision system include the ability to operate in real-time and under a variety of conditions including total darkness. Real-time, processing of 3d data is achieved through a novel implementation of the ICP (Iterative closest point) algorithm. A brief overview of future goals for the on-going project is included.