SBEL is involved in many facets of autonomous vehicle development – from vehicle simulation to reinforcement learning. Autonomous vehicle simulation in Chrono was made possible by two major Chrono modules, Chrono::Vehicle and Chrono::Sensor. More recently, the addition of the SynChrono module has allowed for autonomous vehicle development with tens to hundreds of vehicles in a single simulation. SynChrono uses MPI or DDS-based communication to synchronize the state of multiple vehicles, that then coexist in a time- and space-coherent world.
During the summer of 2020, SynChrono was expanded to support synchronization of deformable terrain, enabling studies of off-road convoys that use reinforcement learning to maintain a convoy as it navigates obstacle fields. This reinforcement learning work leaned heavily on PyChrono which enables Chrono-based reinforcement learning.
In the on-road scene, SBEL has several ongoing projects related to interfacing Chrono’s vehicle simulation with existing physical simulation platforms to enable research at the intersection of human and autonomous vehicle interaction. In an NSF-funded collaboration with two other UW–Madison labs aims to investigate the interplay between human and autonomous drivers – in particular the impact a human driver has on traffic flow when they take over control from a vehicle previous being driven autonomously. As part of this project SBEL has integrated SynChrono with a physical driving simulator located at UW–Madison, enabling a human driver in the simulator to navigate in SynChrono, interacting with other autonomous vehicles.
Further from home, the National Advanced Driving Simulator (NADS) at the University of Iowa maintains a large physical simulator, where a human driver inside a vehicle can drive and interact with a simulated world. An ongoing project is working to connect this physical simulator to SynChrono so that a human driver can drive and interact with autonomous vehicles that are generated by SynChrono.