Description: Automotive transportation is currently evolving at an unprecedented rate. Future vehicles operating in a highly assisted and autonomous mode will need the ability to function on any road possible, in a safe, legal and socially acceptable manner and without access to high-precision, precompiled maps. Therefore, at the heart of any autonomous functionality will be the ability of a vehicle to sense its environment, produce a map of static objects in the environment and to track both itself and other dynamic targets within that environment. For autonomous vehicles to be commercially viable it must achieve this high level of situational awareness using only commercially viable sensors. Multi-sensor data fusion offers the ability to greatly reduce the uncertainty of state estimates and estimate physical states which might otherwise be unobservable.
In addition to the challenge of fusing sensors organic to the vehicle, many future infrastructure projects envisage the concept where vehicles will be connected with each other and with sensors in the infrastructure to further improve situational awareness. This presents many challenges where there may be poorly understood correlation, or poor trust in externally shared information, along with constraints in terms of bandwidth and computational capacity. Furthermore, there exist both challenges and opportunities for understanding and characterising the interaction between the driver and the vehicle.
Organizers: Daniel Clarke, Michael Fiegert, Feihu Zhang, Dhiraj Gulati, Benjamin Noack, Florian Faion