Safe and Effective Collaboration With a High-Payload Robot: A Framework Integrating Novel Hardware and Software Modules

Human-robot collaboration has demonstrated its potential to increase efficiency and productivity in industry. However, inherently safe collaborative robots offer limited payloads and are unsuitable for many heavy lifting and manipulation tasks that would benefit from collaboration. The use of high-payload robots in physical human-robot interaction poses significant safety and performance issues due to the robots' high inertia and power. As a result, these robots either operate autonomously, spatially separated from humans, or at very low speeds, making collaboration ineffective. In this work, we develop an integrated system that allows humans to interact with a high-payload industrial robot in a safe and effective manner. By combining sensorized skin and perception systems to distinguish voluntary from involuntary contacts with a novel variable admittance control method, we achieve safe, stable, and efficient physical human-robot interaction. Our approach is experimentally validated by manipulating a more than 100-kg payload in a real-world scenario of car windshield inspection and preassembly from the automotive industry. In the validation experiments, we achieve faster, more flexible, more efficient, and more human-friendly execution of this task. Our work also contributes on a broader scale by paving the way for the adoption of human-robot comanipulation with high-payload robots.