April 2018
Intermediate to advanced
334 pages
10h 18m
English
One big reason we have already discussed is the variability in vehicle maneuvers that cannot be learned in a supervised manner. In this section, we will go into the details of why reinforcement learning is the best suitable candidate for autonomous driving.
In terms of ML, driving is a multi-agent interaction problem. Consider human drivers driving in a lane without any other cars in proximity. This is way easier compared to changing lanes when there is heavy traffic. The reason why the second scenario is difficult is because it also includes the uncertain and unknown behavior of other drivers. Thus, the number of vehicles interacting with your vehicle, the type of vehicle (small or big), and the behavior of their ...