Human AV Interaction

Indian Institute of Science, Bangalore

*Indicates Equal Contribution
Code

DESCRIPTION

The video demonstrates a study on an Unmanned Ground Vehicle (UGV) navigating through an intersection, equipped with person and lane detection models. As the UGV approaches the intersection, it utilizes a person detection model in combination with a proximity sensor to identify pedestrian. Upon detecting the pedestrian, the UGV halts at a safe distance of 60 centimeters (scaled to 2 meters in the virtual world). This detection and stopping mechanism is a result of the fusion between the proximity sensor data and the detection model. Simultaneously, the lane detection model generates control inputs for the Stanley Controller, which ensures the UGV maintains smooth path tracking between detected lanes. Real-time wheel odometry data is transmitted to the virtual simulation of the autonomous vehicle, where an avatar mimicking a person, equipped with a head-mounted display (Oculus), crosses the road. When the UGV detects a person within the specified threshold, it stops and signals the pedestrian that it is safe to cross the road.