Abstract
The pursuit of long-term autonomy mandates that robotic agents mustcontinuously adapt to their changing environments and learn to solve new tasks.Continual learning seeks to overcome the challenge of catastrophic forgetting,where learning to solve new tasks causes a model to forget previously learntinformation. Prior-based continual learning methods are appealing for roboticapplications as they are space efficient and typically do not increase incomputational complexity as the number of tasks grows. Despite these desirableproperties, prior-based approaches typically fail on important benchmarks andconsequently are limited in their potential applications compared to theirmemory-based counterparts. We introduce Bayesian adaptive moment regularization(BAdam), a novel prior-based method that better constrains parameter growth,leading to lower catastrophic forgetting. Our method boasts a range ofdesirable properties for robotic applications such as being lightweight andtask label-free, converging quickly, and offering calibrated uncertainty thatis important for safe real-world deployment. Results show that BAdam achievesstate-of-the-art performance for prior-based methods on challengingsingle-headed class-incremental experiments such as Split MNIST and SplitFashionMNIST, and does so without relying on task labels or discrete taskboundaries.