Continual Learning for Robotics: Definition, Framework, Learning Strategies, Opportunities and Challenges

Continual learning (CL) is a particular machine learning paradigm where the distribution and learning objective changes through time, or where all the data and objective criteria are never available at once. The evolution the learning process is modeled by a sequence of learning experiences where goal is to be able to learn new skills all along the sequence without what has been previously learned. Continual learning also aims at same time at optimizing the memory, the computation power and the speed the learning process. An important challenge for machine learning is not necessarily finding that work in the real world but rather finding stable algorithms that learn in real world. Hence, the ideal approach would be tackling the real in a embodied platform: an autonomous agent. Continual learning would be effective in an autonomous agent or robot, which would learn through time about the external world, and incrementally develop a of complex skills and knowledge. Robotic agents have to learn to adapt and interact with their environment a continuous stream of observations. Some recent approaches aim at continual learning for robotics, but most recent papers on continual only experiment approaches in simulation or with static datasets. , the evaluation of those algorithms does not provide insights on their solutions may help continual learning in the context of robotics. paper aims at reviewing the existing state of the art of continual, summarizing existing benchmarks and metrics, and proposing a for presenting and evaluating both robotics and non robotics in a way that makes transfer between both fields easier.