@inproceedings {pub4957,
	title = {Fast online IDM parameter estimation for trajectory prediction },
	author = {Karsten Kreutz AND Julian Eggert},
	year = {2022},
	month = {June},
	abstract = {In this paper, we propose and analyze a method for trajectory prediction in longitudinal car-following scenarios. Hereby the prediction is realized by a longitudinal car-following model (Intelligent driver model, IDM) with online estimated parameters. In previous approaches, it has been shown that online parameter adaptation of the IDM is possible but difficult and slow due to the nonlinearity of the parameters, providing only a marginal improvement of prediction quality over the CV, CA baseline models.
 
In our approach, we use the similarity between the parameter-specific trajectory and the real past trajectory as optimization measure. Previously, the IDM parameters were optimized {\textquotedblleft}directly{\textquotedblright}. In the presented approach, we gain the parameters indirectly from a weighted sum of reasonable prototype parameters and optimize these weights. 
 
To show the benefits of the method, we evaluate the properties of our approach against state-of-the-art prediction methods for longitudinal driving, such as CV, CA and particle filter methods on open freeway driving datasets. The evaluation shows that our method provides significant improvements in several aspects: First of all the prediction accuracy is significantly increased, second, the gained parameters exhibit a fast convergence and increased temporal stability and third, the computational effort is reduced so that an online parameter adaptation becomes feasible. 
},
	publisher = {IEEE},
	booktitle = {Intelligent Vehicles 2022},
	editor = {IEEE}
}