A while back someone asked how to star an independent project without having the right technical skills and most of the answers went towards the idea of sharing the project in Slashdot so that someone might pick it up.
Taking that answer and picking up from a post I can't really find about the impact of autonomous driving vehicles on modern societies, I came up with the idea that, being the cost and effort of creating the mobile testing rig (aka, car with a lot of sensors)the major hurdle to a worldwide focused approach to the problem, the creation of a suitable simulation platform would allow for almost anyone to build the model of a car, place some sensors and program an AI based on the output of those sensors.
The concept, therefore, is to create something similar to a car game where the objects would have to be rendered for various wave types and wave lengths, so that we could emulate a surfaces' response to a sonar, lidar ou camera (IR, visible spectrum and UV) and where some sort of scripting language would allow to create an AI agent that would drive the car based on the outputs of the sensors.
There. The idea is out. I am sure that if the idea is without merit I will be crucified by the comments, which is good and will point out where my reasoning has failed. But if it's a good idea, perhaps the open source community would solve on it's own a heck of a challenge.