Meta-Learning-Based Safety-Critical Control in Multi-Obstacles Environments

Autonomous robots operating in diverse scenarios are expected to safely and efficiently adapt to new, unknown, and cluttered environments. In this paper, we introduce a real-time goal-seeking and exploration framework incorporating novel meta-signed distance functions (MetaSDFs) and meta-buffer robust control barrier functions (Meta-BRCBFs). To adapt to environmental changes in real time, we employ Bayesian meta-learning to construct MetaSDFs. Deep neural network weights are initially trained offline, followed by efficient online adaptation at the last Bayesian layer, allowing for online updates at linear time complexity. Each MetaSDF is individually trained for its corresponding obstacle class, enhancing online distance estimation accuracy. Subsequently, buffer zones are constructed around the MetaSDFs to establish corresponding Meta-BRCBFs. These Meta-BRCBFs are activated only when the robot enters these zones, substantially reducing the number of CBFs required. Outside these specified buffer zones, the robot remains in goal-seeking mode, focusing on task completion. After entering a buffer zone, it transitions to exploration mode, prioritizing safety and exploring safe pathways, effectively balancing task execution with environmental adaptability. We demonstrate that, under this framework, the system achieves both safety and asymptotic stabilization. Extensive simulations and experiments are conducted to demonstrate our framework’s effectiveness in both simulated scenarios and real-world environments. These tests confirm our framework’s real-time capabilities and safety assurances in dynamic settings where state-of-the-art methods fail. The video is available at: https://www.youtube.com/watch?v=C6eshldAMxA Note to Practitioners—This paper is motivated by enhancing safety for autonomous robots operating in diverse and changing environments. Existing methods, which utilize traditional or incremental neural networks, fail to perceive environmental changes and obtain the nearest distances to obstacles for safety in real time. Additionally, with the increased number of obstacles, the real-time performance of optimization solvers for such approaches is further reduced, which quickly causes error motions. To overcome these challenges, this paper introduces a novel approach enabling online updates of the trained neural network so that it adapts to unknown environments rapidly. Buffer zones are established around obstacles, allowing us to focus solely on those that the robot enters into these zones. This eliminates the need to consider all obstacles simultaneously, significantly reducing the time required for the optimization solver. The safety and asymptotic stability have been rigorously proven through mathematical validations. Extensive simulations and physical experiments demonstrate the framework’s real-time and safety performance in complex and variable scenarios. Future research will extend this approach to robots with higher-dimensional control systems, such as robotic manipulators.