Autonomous Driving:
We just get from the sensors (Lidar, Radar, Video) the position of the other objects (x, y) and directly (Radar x´) and indirectly (LAser, Video by differentiating y we get y´).
How can we get x´´ and y´´, which means the prediction of the trajectory of the other objects?
1.) We can extrapolate x´and y´ by Kalman filters.
2.) We can make a driving dynamics prediction by taking Kamm´s circle.
3.) We can make a prediction based on the environment like curves and stop signs, traffic lights.
4.) We make an ethical prediction by estimating how the other driver/pedestrian/cyclist behaves to not cause an accident.
Especially to points 3 and 4 I would like to get bettern known the state of the art. Who knows the relevant research in this field?
Thanks
With any predictive model, human behavior is tricky and one cannot really, that I know of, predict it with any high degree of certitude. To start, you would need an enormous volume of watched data, be it driver or pedestrian movements.
That caution said, we are in the business of data gathering and observations, so who better than a seasoned traffic engineer to make learned predictions and then to inform ourselves or audiences that these are our best prognostications of how vehicles or people move.