TORC Robotics has been subcontracted through the Robotics Technology Consortium (RTC) to develop an advanced sensor fusion system for the Department of Defense that will significantly increase high-speed obstacle detection range. This long-range obstacle detection, classification and prediction system will enhance autonomous navigation capabilities for unmanned ground vehicles operating in mission-relevant environments at speeds up to 100 KPH. The system will be capable of detecting and maintaining a variety of tracking statistics for each obstacle. TORC will incorporate these enhanced capabilities with its core autonomy framework for future availability in its AutonoNav™ product line.
To support integration with the project sponsor’s autonomy framework, TORC will develop an Application Programming Interface (API) for the advanced sensor fusion software, and build a hardware prototype capable of installation on a range of vehicles including the HMMWV and LMTV. The system fuses asynchronous and heterogeneous sensor modalities through a joint probabilistic data association approach to reduce false positive/negative data, which is essential to high-speed autonomous navigation. TORC will achieve long-range detection and classification through a combination of commercial-off-the-shelf LIDAR, radar and vision technologies from Ibeo, Velodyne, and Smartmicro. TORC will assess sensor and fusion performance at high-speed under a variety of man-made weather conditions to include rain, dense fog and snow at the Virginia Tech Transportation Institute (VTTI) Smart Road.
“Despite recent deployments of full-sized autonomous vehicles for operational assessments in Afghanistan, most are still quite limited in their ability to operate autonomously and at high speed outside of their pre-planned scenarios,” states Andrew Culhane, Business Development Manager at TORC. “The reality is that military UGVs need to be able to operate autonomously within complex mission environments while keeping pace with the force. In order for that to happen, UGV perception technology must be capable of detecting, classifying and predicting obstacles at longer ranges while moving at operational speeds.”
TORC will leverage software initially conceptualized under a DARPA SBIR now under further research on the RTC Sensor-based Collision Prediction project. This project provides the software architecture and obstacle prediction capabilities that TORC will extend to meet the requirements for this project. The primary research platform for this project is the ByWire XGV™, a drive-by-wire controlled ground robotics vehicle.
TORC will demonstrate the new long-range obstacle detection, classification and prediction capabilities in 2012. Results of this work should open up new possibilities for deploying autonomous vehicles for missions that demand navigating at higher speeds in unstructured environments