Safe Robot Driving in Cluttered Environments

The Navlab group at Carnegie Mellon University has a long history of development of automated vehicles and intelligent systems for driver assistance. The earlier work of the group concentrated on road following, cross-country driving, and obstacle detection. The new focus is on short-range sensing, to look all around the vehicle for safe driving. The current system uses video sensing, laser rangefinders, a novel light-stripe rangefinder, software to process each sensor individually, and a map-based fusion system. The complete system has been demonstrated on the Navlab 11 vehicle for monitoring the environment of a vehicle driving through a cluttered urban environment, detecting and tracking fixed objects, moving objects, pedestrians, curbs, and roads.* 1 The Need For 360 Degree Safeguarding Robot driving has concentrated on forward-looking sensing, for road following and obstacle detection. This is an appropriate first step, but real deployment of mobile robots will require additional sensing and reasoning to surround the robot with safeguard sensors and systems. Our group is currently building short-range sensing to surround vehicles to improve the safety of robotic and human-controlled vehicles. In the civilian context, our focus is driver assistance for transit busses. Busses drive at relatively slow speeds, in crowded urban environments. One of the most frequent types of accidents in transit busses is side collision, where the driver does not have adequate awareness of objects near the bus, then turns too sharply and sideswipes a fixed object or (less often) hits a pedestrian. Preventing these accidents requires shortrange sensing along the side and front of the bus, detecting fixed objects, detecting and tracking moving objects, predicting the motion of the bus itself, and a suitable driver interface for alerting the driver. In the military context, our focus is short-range sensing for full automation of scout vehicles. An autonomous vehicle moving through a cluttered environment, such as a forest, may need to move between objects (e.g. trees) with very little clearance on either side of the vehicle. The conventional approach is to sense trees with forwardlooking sensors, enter those trees into a map, and estimate the clearance as the vehicle moves forward and the trees move out of the field of view of the forward-looking * Prepared through collaborative participation in the Robotics Consortium sponsored by the US Army Research Laboratory under the Collaborative Technology Alliance Program, Cooperative Agreement DAAD19-01-2-0012. The US Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation thereon. Additional funding from the Federal Transit Administration, US DOT; from Bosch Corporation; and from SAIC Corporation. sensor. If sensor data is noisy, or if the vehicle slips and slides in mud, the estimated clearance may be incorrect. It is better to directly and continuously sense nearby objects all along the side of the vehicle as it moves through the forest. Both civilian and military vehicles, and both driver assistance and full automation, need to pay special attention to moving objects, and particularly to humans. People move in unpredictable ways. Seeing a person with a forward-looking sensor, or having the driver note the position of a person in front of a bus, is no guarantee that the person will remain safely out of the vehicle’s way. Thus, we pay special attention to detecting moving objects, classifying them if possible as people, continuously tracking their motion, and attempting to predict their future actions. This work is part of a trend in the intelligent vehicle research community, focusing on better sensing for driver assistance. The Intelligent Vehicles 2002 Conference, for instance, included a number of papers on pedestrian detection 2 3 4 5 6 and on urban driving 7 .