Nicolas ALBERTO TORRES will defend his thesis on October 23, 2023 at 10:30 a.m., (Salle Ada Lovelace (2eme etage), Inria Universite de Bordeaux)on the subject:”Predictive Control of Collaborative Robots in Dynamic Contexts”.
Collaborative robotics involves the transformation of industrial robots to function in shared workspaces along-side humans, resulting in more compact and manageable robotic systems. This evolution necessitates a reevaluation of the requirements for robot controllers. Safety remains paramount, but robots must also adapt to dynamic contexts where objects are in constant motion, and conditions are ever-changing as people carry out their tasks.To address these challenges, this work proposes using Model Predictive Control (MPC) to empower the robot to dynamically adjust its operations in a responsive manner. This enables the implementation of human-aware behaviors that prioritize safety, as detailed in . The methodology, based on linear MPC, primarily focuses on precise position and orientation control, leveraging a mathematical formalism outlined in a conference paper .
This control architecture serves as a versatile framework capable of accommodating complex tasks that extend beyond simple point-to-point movements in space, as discussed in . Through real-time, high-frequency re-planning, this work showcases the robot’s agility in responding to evolving tasks and dynamic environments.
Furthermore, this research addresses the issue of using the robot’s full range of motion capabilities. In contrast to prevailing practices that rely on simplified Cartesian limits for trajectory planning, this work proposes a method for utilizing the robot’s actual actuator capabilities by mapping joint-space capacities onto the operational workspace.
This enhancement can seamlessly integrate with the MPC-based architecture, facilitating rapid re-planning that simultaneously considers user-defined workspace constraints and the robot’s true capabilities.
If you have a request or questions about the laboratory, please contact our team.