A Cognitively Informed Model for Perception in Driving
Publised at IROS 2019, Workshop: “Towards Cognitive Vehicles: perception, learning and decision making under real-world constraints. Is bio-inspiration helpful?”
Deep learning is responsible for the current renewed success of artiﬁcial intelligence. Applications that in the recent past were considered beyond imagination, now appear to be feasible. The best example is autonomous driving. However, despite the growing research aimed at implementing autonomous driving, no artiﬁcial intelligence can claim to have reached or closely approached the driving performance of humans, yet. Deep learning is an evolution of artiﬁcial neural networks introduced in the ’80s with the Parallel Distributed Processing (PDP) project . There is a fundamental difference in aims between the ﬁrst generation of artiﬁcial neural networks and deep neural models. The former was motivated primarily by the exploration of cognition. Current deep neural models are instead developed with engineering goals in mind, without any ambition or interest in exploring cognition. Some important components of deep learning – for example reinforcement learning or recurrent networks – owe indeed an inspiration to neuroscience and cognitive science, as PDP far legacy. But this connection is now neglected, what matters is only the pragmatic success in applications. We argue that it urges to reconnect artiﬁcial modeling with an updated knowledge of how complex tasks are realized by the human mind and brain. In this paper, we will ﬁrst try to distill concepts within neuroscience and cognitive science relevant for the driving behavior. Then, we will identify possible algorithmic counterparts of such concepts, and ﬁnally build an artiﬁcial neural model exploiting these components for the visual perception task of an autonomous vehicle.