The growing use of robots in safety-critical settings creates a pressing need for control methods that remain reliable under model errors, changing conditions, and uncertain interactions. Learning-based control can improve adaptability and task performance by extracting useful control information from data. However, its deployment is complicated by distribution shifts, incomplete safety assurances, and limited robustness outside nominal operating conditions. This workshop will examine how learning-based methods can be combined with safety- and resilience-oriented control principles to address these limitations. The program will span foundational theory, algorithmic development, and experimental validation, with particular attention to preserving learning flexibility without compromising rigorous safety requirements. Invited speakers will highlight recent advances and demonstrate how these methods have been translated from theoretical frameworks into functioning robotic platforms.

As autonomous systems are increasingly integrated into safety-critical applications, ensuring their reliable performance under uncertainty is crucial. Traditional control methods often face difficulties balancing high performance with risk management, especially in dynamic environments with incomplete or evolving data. This workshop will address these challenges by exploring the integration of data-driven approaches with risk-aware control strategies. It will feature both theoretical advancements and real-world applications, focusing on balancing the uncertainty inherent in data-driven methods with the conservatism required for risk-aware control. Invited speakers will present state-of-the-art advancements, progressing from theoretical frameworks to successful implementations in autonomous systems. Case studies from domains such as autonomous vehicles, medical robots, and industrial processes will demonstrate how risk-aware control can enhance system safety and reliability in unpredictable environments.