Biological Intelligence arises from the tight interaction between the brain and the body, that co-evolved to optimise the resources needed for perception and action and their results. Part of the computation is performed by the body and sensors and the way information is acquired shapes how computation is performed. Efficient and robust artificial intelligence can emerge from similar processes, supporting the development of Embodied Neuromorphic Intelligence, that studies the synergies between body and computation.
Event-driven sensors generate information only with movement, static stimuli can only be perceived upon exploratory actions. We transfer visual attention models to their event-driven, spiking, counterpart, to equip iCub with a low-latency selection of relevant regions for saccades and computationally intensive inspection. Tactile exploration will follow the exploratory procedures of humans, guided by event-driven proprioception and tactile information, using unique neuromorphic multi-modal tactile sensors. A coherent representation of vision, touch and proprioception for the robot's body will be generated through self-exploration. Finally, we are exploring capacitive, resistive, and piezoelectric materials, and mixed-mode subthreshold CMOS circuits with the aim of combining them for a hybrid skin capable of encoding efficiently a broad range of tactile stimuli with low latency and low power consumption. We are expanding this research to soft robots, starting from tactile sensing and its use in control. The same technologies can be used to interface prosthetic devices, using EMG for control and touch for sensory feedback.
Methods: biologically inspired models of vision (attention, depth, and motion perception and stabilisation), touch (encoding, exploratory procedures) and robot's body schema implemented using Spiking Neural Networks and spike-driven learning. Subthreshold CMOS circuit design for emulation of human glabrous skin tactile afferents on silicon and flexible substrates. Asynchronous AER serial communication for robots and distributed systems.
