SoldierVision: Visual Measurement Based Autonomous Navigation

Primordial, Evolution Robotics and Professor Clark Olson propose SoldierVisionTM: a vision-based autonomous navigation system for soldiers in the GPS-denied battlefield. Existing wearable dead-reckoning devices, relying on inertial measurement units (IMUs) and stride estimates, rapidly lose accuracy as the soldier moves and are currently of limited value for the dismounted soldier. However, recent advances in computer vision algorithms have opened up the possibility of using low cost cameras (e.g. mounted to the soldier’s helmet) to perform vision-based localization and navigation. The proposed SoldierVision product will build on our partner’s (Evolution Robotics’) state-of-the-art vSLAMTM technology, which relies on pattern recognition to automatically create a map of the visible surroundings and localize the camera within the environment. Evolution Robotics developed the pattern recognition software for the Sony Aibo robot and has collaborated with the military on projects ranging from RPG detection to UAV autonomous navigation. An impressive recent application of computer vision techniques to localization and navigation was the use of binocular camera-based odometry to assist the Mars Rovers in their semi-autonomous exploration of the Martian surface (a truly GPS-denied environment). At the Jet Propulsion Laboratory, our team member Professor Clark Olson spent 5 years working on computer vision techniques for the Mars rovers, and we are excited to leverage his expertise in this area. Also, Primordial is in an excellent position to commercialize this product, having already integrated our patented Ground Guidance® navigation software into the Army’s Land Warrior system for dismounted soldiers. In phase I, Primordial will develop a prototype version of SoldierVision, building on Evolution Robotics’ innovative vSLAM technology, and we will perform a live demo with human subjects in order to realistically assess the feasibility of the proposed system.