Zero-shot Safety Prediction for Autonomous Robots with Foundation World Models

arXiv:2404.00462v1 Announce Type: new
Abstract: A world model creates a surrogate world to train a controller and predict safety violations by learning the internal dynamic model of systems. However, the existing world models rely solely on statistical learning of how observations change in response to actions, lacking precise quantification of how accurate the surrogate dynamics are, which poses a significant challenge in safety-critical systems. To address this challenge, we propose foundation world models that embed observations into meaningful and causally …

abstract arxiv autonomous autonomous robots change cs.lg cs.ro dynamic foundation however prediction quantification robots safety statistical systems train type world world model world models zero-shot