Although global pose is essential for success in certain vehicle missions and tasks, when GPS is denied there is no single technology (source) that can robustly estimate global pose without suffering from one or more failure modes. We propose to design an architecture that provides robustness by integrating automated celestial navigation, registering on-board sensor data to a map (map registration) or to sensor data from a previous traverse (route registration or place recognition), wheel odometry, measuring vehicle motion from camera data (visual odometry) or range data (range odometry), and inertial measurements of acceleration and angular rates (IMU) to compensate for all possible failure modes. The architecture will provide a direct alternative to GPS over the widest range of conditions possible. The architecture will be built around Carnegie Robotics LLCs (CRLs) SmoothPose product. The proposed system supports autonomous on- and off-road driving, under tree canopy, amongst tall buildings, in deep ravines, on slippery surfaces, in cluttered and open terrainall with GPS signals denied.
Benefit: Robust positioning is a must have 0x9D for any unmanned system that can operate with some degree of autonomy in traditional field robotics environments. Its importance as a core robotics component ranks as high as, or even higher than robot perception. Quite simply a robot that doesnt know its position cannot follow a path, avoid an obstacle, or reach a delivery point. GPS is the workhorse global positioning modality; however it fails due to jamming and degrades under canopy, in ravines, amongst tall buildings and in tunnels. As a core robotics component, the enhanced SmoothPose product would be sought after by OEMs who wish to field the next generation of intelligent vehicles to their end users in mining, agriculture and material handling markets.
Keywords: Autonomous Unmanned Ground Vehicle, Autonomous Unmanned Ground Vehicle, Particle Filter, Kalman Filter, GPS-denied global pose,
