Simulating Crowds for Traffic Safety Research

From ISLAB/CAISR
Title Simulating Crowds for Traffic Safety Research
Summary Integrate crowd simulation into a mixed-reality platform for development and testing of advanced automotive safety systems.
Keywords Virtual world, swarm simulation
TimeFrame
References http://gamma.cs.unc.edu/research/crowds/

http://www.coppeliarobotics.com/

Prerequisites Solid programming in Python, C, or C++; previous experience with ROS and mobile robotics would be a significant advantage
Author
Supervisor Roland Philippsen
Level Master
Status Internal Draft

Generate PDF template

Project Description

Many (if not most) safety-relevant elements of city driving involve vulnerable road users: pedestrians, bicycles, or kids playing in the street, to name just a few. Advanced safety systems for cars and trucks should take these into account, and there are high-profile research projects that specifically investigate how best to protect these vulnerable road users from accidents. One possibility is to incorporate people detection algorithms into cars, based on onboard sensors, and provide emergency avoidance maneuvers in case of imminent collisions.

Such active safety systems would need extensive testing before being admitted in series production cars and trucks. But testing them is quite challenging: because of the severity of system failures, it is not possible to test them on real pedestrians. This raises the question of how to emulate or simulate pedestrians, for instance with puppets or in mixed reality settings. A large body of prior work exists in simulating crowds of humans for applications in computer games or the analysis of human movement through confined spaces such as subway stations. Some examples can be found on the website of the GAMMA research group at the University of North Carolina at Chapel Hill.

In this project, the student will survey the state of the art in crowd simulation, with a focus on approaches and methods that can be used in mixed-reality settings to develop, evaluate, and test active safety systems in intelligent vehicles. Then, two or three promising algorithms will be implemented within our existing augmented reality robot platform, which then allows to simulate the sensor data of virtual vehicles that are confronted with these crowds of humans. The most appropriate crowd simulator will then be used to implement a handful of very specific test scenarios for active safety systems, to be further defined in collaboration with our industrial partner.

Further References