SafeDMPs: Integrating Formal Safety with DMPs for Adaptive HRI

Soumyodipta Nath, Pranav Tiwari, Ravi Prakash

Keywords: Spatio Temporal Tubes, Dynamic Movement Primitives, Human Robot Interaction, Reactive Task Transition, Formal Safety

TL;DR: A modular planning and control framework that combines Dynamic Movement Primitives (DMPs) for smooth, generalizable trajectory generation with Spatio Temporal Tubes (STTs) for safety enforcement

Abstract:

Robots operating in human-centered environments must generate motions that are not only adaptive and responsive, but also provably safe in real time. While temporal logic-based planners enable structured high-level task specification, executing these plans safely at the motion level remains challenging. Existing approaches based on Control Barrier Functions (CBFs) guarantee safety but incur significant computational overhead due to online constrained optimization.
We propose a modular planning and control framework that combines Dynamic Movement Primitives (DMPs) for smooth, generalizable trajectory generation with Spatio-Temporal Tubes (STTs) for safety enforcement. Unlike CBFs, STTs avoid online optimization and provide closed-form feedback laws, ensuring real-time, collision-free execution.
We validate our approach on a Franka Emika robot performing collaborative tasks such as whiteboard writing and adaptive recovery under human intervention. Compared to CBF and Neural ODE baselines, our method achieves up to 99.97% faster execution, and 48% lower memory usage, while maintaining formal safety guarantees

Submission Number: 19