Local sensorimotor control and learning in robotics with organic neuromorphic electronics

Artificial intelligence applications have demonstrated their enormous potential for complex processing over the last decade, however they still lack the efficiency and computing capacity of the brain. In living organisms, data signals are represented by sensory and motor processes that are distributed, locally merged and capable of forming dynamic sensorimotor associations through volatile and non-volatile connections. Using similar computational primitives, neuromorphic circuits offer a new way of intelligent information processing that makes it possible to adaptively oberserve, anaylze, operate and interact in real-world scenarios [1-6]. In this work we present a small-scale, locally-trained organic neuromorphic circuit for sensorimotor control and learning, on a robot navigating inside a maze. By connecting the neuromorphic circuit directly to environmental stimuli through sensor signals, the robot is able to respond adaptively to sensory cues and consequently forms a behavioral association to follow the way to the exit. The on-chip sensorimotor integration with low-voltage organic neuromorphic electronics opens the way towards stand-alone, brain-inspired circuitry in autonomous and intelligent robotics.