Mapping Of Local Environments for Reconnaissance And Training (MOLE-RAT)

Unmanned ground systems provide the warfighter with a variety of standoff capabilities. One application is remote entry and exploration of urban structures and other potentially dangerous enclosed spaces. A robot capable of autonomously navigating these structures while simultaneously building a 3D map of the interior would provide soldiers with data for interactive reconnaissance and mission rehearsal. We propose a solution that uses simultaneous localization and mapping (SLAM) to track the robot’s location and build a 3D model of the scene using a stereo video camera and standard onboard motion sensors such as IMU and odometry. A particle filter is used to efficiently model and update the robot position and scene model, allowing the system to run in real-time. To mitigate the lack of scene texture typical of indoor imagery, we will employ line features in addition to traditional point features. Our system will also apply geometric constraints, such as planar surfaces and parallel walls, to generate more accurate and reliable location and mapping results. As part of the development effort we plan to perform a thorough system evaluation which will assess the effectiveness of system components and guide the construction of a full system prototype.