Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Human-Robot Collaboration in Shared Tasks by Model Predictive Control
Robots that collaborate in direct interaction with humans like prostheses, assistive devices, and rehabilitation robots need to synchronize their behavior with the human's actions. These actions cannot be predicted precisely. Therefore a controller is needed that is robust to this uncertainty.
Keywords: Human-Robot Collaboration, Human-Robot Interaction, Model Predictive Control, MPC, Robust Control, Torque Control, Rehabilitation Engineering, Biomedical Engineering, Health Science
**The Background** Collaborative robots in the industry have a great potential to make human workers more efficient by offering them a helping hand. In these applications the subtasks are usually distributed to the two systems. E.g. the robot picks up and holds PCB's while the human interacts with it. In this case triggers suffice to synchronize the transition from one to the next stage of the task. However, in mechatronic systems that are replacing a function of the human body e.g. prostheses, assistive devices, and rehabilitation robots the human and the machine have to collaborate in the same task. In most of these cases the human is the master in the collaboration. This means that the robot has to continuously react to the actions of the human to be successful in the common task. As the human behavior is only predictable with large uncertainty a robust controller is needed.
**The Project** In this project we will focus on the application for a rehabilitation robot with exoskeleton structure. This device assists and corrects the motion of the impaired arm of patients with neural disorder (e.g. stroke) during therapy sessions. The other arm of such patients is often healthy. Studies have indicated that training the affected arm in bi-manual tasks together with the healthy promotes a fast recovery. For this bi-manual therapy the robot has to collaborate with the healthy arm in the same task.
**Your Task:** We want to start this project of bi-manual therapy by an evaluation in simulation. We will simulate the motion of the healthy arm in a task and predict the motion by a simple algorithm (more advanced estimators will be investigated later). We want to show on some example tasks, that an MPC controller can robustly succeed in the task using this simple motion prediction of the collaborative system and the task description.
**The Background** Collaborative robots in the industry have a great potential to make human workers more efficient by offering them a helping hand. In these applications the subtasks are usually distributed to the two systems. E.g. the robot picks up and holds PCB's while the human interacts with it. In this case triggers suffice to synchronize the transition from one to the next stage of the task. However, in mechatronic systems that are replacing a function of the human body e.g. prostheses, assistive devices, and rehabilitation robots the human and the machine have to collaborate in the same task. In most of these cases the human is the master in the collaboration. This means that the robot has to continuously react to the actions of the human to be successful in the common task. As the human behavior is only predictable with large uncertainty a robust controller is needed.
**The Project** In this project we will focus on the application for a rehabilitation robot with exoskeleton structure. This device assists and corrects the motion of the impaired arm of patients with neural disorder (e.g. stroke) during therapy sessions. The other arm of such patients is often healthy. Studies have indicated that training the affected arm in bi-manual tasks together with the healthy promotes a fast recovery. For this bi-manual therapy the robot has to collaborate with the healthy arm in the same task.
**Your Task:** We want to start this project of bi-manual therapy by an evaluation in simulation. We will simulate the motion of the healthy arm in a task and predict the motion by a simple algorithm (more advanced estimators will be investigated later). We want to show on some example tasks, that an MPC controller can robustly succeed in the task using this simple motion prediction of the collaborative system and the task description.
- literature research
- setting up the simulation environment
- developing a robust MPC controller for this task
- further work packages might be added depending on the progress
- literature research - setting up the simulation environment - developing a robust MPC controller for this task - further work packages might be added depending on the progress
We are searching for motivated students with:
- good understanding of multi-body systems
- good understanding of control system theory
An advantage is:
- working experience in C++
- working experience with ROS
- knowledge of advanced control methods (e.g. lectures like Model Predictive Control, Recursive Estimation, Dynamic Programming and Optimal Control etc.)
We are searching for motivated students with:
- good understanding of multi-body systems - good understanding of control system theory
An advantage is:
- working experience in C++ - working experience with ROS - knowledge of advanced control methods (e.g. lectures like Model Predictive Control, Recursive Estimation, Dynamic Programming and Optimal Control etc.)
If you are interested in the project please apply with your CV, academic transcript, and a short description of your interest in the project to:
Farbod Farshidian (farshidian@ethz.ch) and
Yves Zimmermann (yvesz@ethz.ch)
If you are interested in the project please apply with your CV, academic transcript, and a short description of your interest in the project to: Farbod Farshidian (farshidian@ethz.ch) and Yves Zimmermann (yvesz@ethz.ch)