Personalized Gap Selection Algorithm for On-ramp Merging Based on Virtual Game

On-ramp merging is a challenging task to tackle for driving intelligence for high-level automated driving, in which selecting a suitable gap is vital for safety and efficiency. To take the oncoming dynamic interaction with other vehicles into consideration, a virtual game method including states prediction and level-k gaming is proposed, which can filter gaps inappropriate to merge to in advance. Then a gap selection algorithm based on utility with personalized parameters is used to select the best gap after the exclusion of virtual games. Driver-in-the-loop experiments are carried out to collect human driving data, which is used for personalized parameters calibration and validation. Test cases show that the gap selected by the well-calibrated algorithm is mostly consistent with different human drivers, reflecting the personalization ability.