NOTC with Feature Descriptors Applied To Autonomous Driving: Towards Autonomous Driving Without Any Prior Knowledge

摘要

Autonomous driving is still an open and challenging problem. This is specially true because in a real-world scenario the information from sensors is high-dimensional, sometimes unreliable and only indirectly related with the objective of driving the car. In this paper, we focus on the hard problem of using only images taken from a camera in front of the car to drive it. Moreover, instead of creating algorithms to control the car, a learning algorithm called Novelty Organizing Team of Classifiers (NOTC) is used together with a pre-processing stage. Preliminary results are reported in simulation.