Safe learning is central to AI-enabled robots where a single failure may lead to catastrophic results. Existing safe learning methods are not scalable, inefficient and hard to train, and tend to generate unstable signals under noisy inputs that are challenging to be deployed for robots. To address these challenges, we propose Adaptive explicit-Barrier Net (ABNet) in which barriers explicitly show up in the closed-form model that guarantees safety. The ABNet has the potential to incrementally scale toward larger safe foundation models. Each head of ABNet could learn safe control policies from different features and focuses on specific part of the observation. In this way, we do not need to directly construct a large model for complex tasks, which significantly facilitates the training of the model while ensuring its stable output. Most importantly, we can still formally prove the safety guarantees of the ABNet. We demonstrate the efficiency and strength of ABNet in 2D robot obstacle avoidance, safe robot manipulation, and vision-based end-to-end autonomous driving, with results showing much better robustness and guarantees over existing models.

How to make an AI model/robot learn like a human being is a long-standing challenge. The human-style of learning is in a scalable and reliable way. In other words, a human always safely learns skills one by one instead of learning all the skills at once, which is contradictory to existing large foundation models that require huge data and extremely expensive computation resource. This work marks a first step towards safe and scalable learning using our proposed ABNet. ABNet is efficient, scalable, and most importantly with provable safety guarantees, and it inspires us to further explore learning beyond safety requirements. ABNet will speed up the grounding of AI on real robots, safely and efficiently assisting humans in physical and cognitive tasks.