shivesh.mecheng@gmail.com
The recent surge in high-DoF robotic platforms, from humanoids to mobile manipulators, presents both fundamental challenges and opportunities in whole-body control for integrated locomotion and manipulation. This workshop will bring together leading researchers and practitioners to discuss robust, real-world control frameworks that coordinate body posture, balancing, arm motion, and environment interaction for complex manipulation tasks in unstructured settings.
The workshop focuses on advancing whole-body control methods that are robust and ready for real-world robotic deployment. While recent learning-based techniques have shown impressive progress, they often struggle with reliability, safety guarantees, and generalization in unconstrained environments. Conversely, purely model-based methods can be difficult to scale and struggle with real-world uncertainty. To address these limitations, the workshop will emphasize hybrid approaches that integrate data-driven learning with physics-grounded models to improve reliability, interpretability, and performance. A key goal is to bridge these paradigms, leveraging the adaptability of learning-based methods while retaining the structure, stability, and safety guarantees of model-based control, while discussing open challenges and outlining future directions for the field.
Topics include model-based versus learning-based control, optimization and hierarchical control strategies, imitation learning, whole-body perception and sensing, human-robot interaction, and practical deployment scenarios. Through invited talks, panel discussions, and exchanges between academia and industry, participants will explore emerging benchmarks, challenge assumptions in current methodologies, and identify future directions toward agile, autonomous robotic systems capable of seamless whole-body manipulation in the real world.