Learning Self-Awareness for Autonomous Vehicles: Exploring Multisensory Incremental Models

The technology for autonomous vehicles is close to replacing human drivers by artificial systems endowed with high-level decision-making capabilities. In this regard, must learn about the usual vehicle's behavior predict imminent difficulties before they happen. An agent should be capable of continuously interacting multi-modal dynamic environments while learning unseen novel concepts. Such are not often available train on it, so have an understanding its own capacities and limitations. This usually called self-awareness. paper proposes a self-awareness modeling signals coming from different sources. shows how machine techniques can used under generic framework single modality models using Dynamic Bayesian Networks. presented case, probabilistic switching model bank generative adversarial networks employed positional visual information respectively. Our results include experiments performed real vehicle, highlighting potentiality proposed approach at detecting abnormalities in scenarios.