ORION: Operational Robot with Intelligent Off-road Navigation

For Robotic/Autonomous Systems (RAS) to be truly effective in the battlefield, they need to be integrated with intelligent decision-making capabilities. In particular, the following capabilities will help them dealing with the challenging real-world problems of off-road navigation: 1) Accurate terrain traversability assessment; 2) Optimal trajectory computation; and 3) Effective maneuver selection. Understanding how to implement these intelligent capabilities is the key to elevate a robot to become a trustworthy partner to human warfighters and can strengthen manned-unmanned teaming (MUM-T). In this STTR effort, we have been developing ORION, a computational framework that implements such intelligent capabilities—treating traversability assessment as a classification problem, optimal trajectory computation as a heuristic motion planning problem, and maneuver selection as a policy mapping problem. In this effort, those problems are being addressed by employing state-of-the-art machine learning and artificial intelligence techniques such as the convolutional neural networks and sampling-based kinodynamic planning. In Phase I, by prototyping and field-testing the key/high-risk capabilities using an operational robot, the development risks in Phase II were successfully mitigated. The overall goal of ORION Phase II is to develop the second generation of the ORION prototypes (ORION-2) that can demonstrate its full capabilities in relevant environment.