Victor BOUTIN - Early-Stage Researcher - 2016 Class

The Ph.D. project aims to control an aerial robot (drone) with human gestures using bio-inspired neural networks. The main originality of the project is to achieve such control by taking advantage of the distributed, dynamical representation of information in the brain which is leading to anticipatory and predictive behavior. Thus, the scientific challenge will result in developing and implementing spiking neural networks and learning rules involved in low level vision (from the retina to cortical areas V1 and V4), which are relevant for gesture recognition tasks. This will necessitate to organize the networks in a hierarchical model, and to develop a bio-plausible unsupervised learning rule based on energy minimization and sparseness maximization. The goal is to achieve performances (in term of response time, accuracy, speed of training…) close to its biological counterpart.