(almost identical to what reference driver did)
(managed to do quickly correct its trajectory in the very last moment)
(but agent also failed to use that huge gap at the end)
(Seemingly it would take too long because of driving at ~60 km/h, reward pressure)
(also a common example of speeding up just before exiting can be observed)
(better reflex than reference driver)
(weird behavior of human driver, edge case)
(could end up well, but acceleration was not strong enough)