Automated Crossing of Intersections Controlled by Traffic Lights

The crossing of intersections controlled by traffic lights requires decision making: at the onset of yellow, drivers must quickly choose whether to stop or to pass and, in the latter case, with which speed. Sometimes drivers select the incorrect action, passing too late or performing sudden decelerations. This paper describes a driver support system for traffic light negotiation, which is based on a layered control architecture with human-like motor primitives. The system receives information regarding the traffic light timings, produces a set of maneuvers every 100 ms and then selects the motion with the minimum initial jerk. The stop maneuver has been validated using real data collected from human drivers. The developed system may be integrated, within subsumption architectures, to produce an artificial driver for autonomous driving applications in urban scenarios.