Neurocognitive–Inspired Approach for Visual Perception in Autonomous Driving

Since the last decades, deep neural models have been pushing forward the frontiers of artificial intelligence. Applications that in the recent past were considered no more than utopian dreams, now appear to be feasible. The best example is autonomous driving. 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. While the early forms of artificial neural networks were aimed at simulating and understanding human cognition, contemporary deep neural networks are totally indifferent to cognitive studies, they are designed with pure engineering goals in mind. Several scholars, we included, 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. More specifically, we will point to four neurocognitive theories: the simulation theory of cognition; the Convergence–divergence Zones hypothesis; the transformational abstraction hypothesis; the free–energy predictive theory. Our proposed model tries to combine a number of existing algorithms that most closely resonate with the assumptions of these four neurocognitive theories.