Motion Planning and Interactive Decision Making

Motion Planning and Interactive Decision Making

The push for the deployment of autonomous vehicles (AV) is already a reality in many cities and roads around the world. AVs make use of the improved capabilities in sensing and perception to predict the motion of other agents (human-driven vehicles, cyclists, and pedestrians) and plan a safe trajectory accordingly. Self-driving cars can help diminish the amount of accidents because of their faster reaction times and lack of distraction; however, they still cannot handle all types of traffic scenarios. Some of the most problematic situations to tackle are maneuvers in dense traffic conditions. Humans use a combination of visual cues like blinkers or hand signs and actions such as nudging, braking/accelerating to open/close a gap in order to communicate their intentions to surrounding agents.

Our Interactive Decision Making work at HRI includes modeling and simulating the behavior of surrounding agents to safely coordinate the combined actions. By creating motion planning behaviors that proactively seek information about others’ intentions while at the same time attempting to convince other human-driven vehicles to accommodate the AV’s desired maneuver, our system is able to prevent deadlocks.