Extending a whole-body control framework for manipulator motion behavior generation to dynamic tasks
A stack of tasks formulation for whole-body robot motion control allows for imposing a hierarchy of desirable properties on the output behavior of the system. Tasks specified at higher priority levels are satisfied first, before moving on to refine solutions further and satisfy tasks of lower priority. Typically, a high-priority task is used to ensure that the robot does not collide with objects in the environment (or itself).
The main goal of this thesis is to extend an existing framework for kinematic motion control (which is based on embedded optimization [1]) to dynamic control allowing for force-based task formulations [2]. Evaluation will be done in simulation using the MuJoCo simulator [3] and a model of a dual-arm robot (ABB's YuMi). As application example assisted teleoperation for manipulation is envisioned – i. e., a human operator provides wrist pose commands via a magnetic pose tracker, while the control scheme is responsible for auxiliary tasks such as obstacle avoidance and limiting interaction forces with the environment. If there is time left, the thesis can include an experimental evaluation of the teleoperation framework using an independent subject group.
Tasks:
- Extending an existing kinematic control framework to dynamic control using the ros_control framework
- Defining and implementing motion behaviors for assisted teleoperation of manipulation tasks
- Integrating a simulator tailored to interaction tasks [3] into the ROS ecosystem
Necessary skills:
- Solid C++ knowledge
- Some background in multi-body dynamics
- Experience in the ROS ecosystem would be helpful
[1] Kanoun, Oussama, Florent Lamiraux, and Pierre-Brice Wieber. "Kinematic control of redundant manipulators: Generalizing the task-priority framework to inequality task." IEEE Transactions on Robotics 27.4 (2011): 785-792.
[2] Saab, Layale, et al. "Dynamic whole-body motion generation under rigid contacts and other unilateral constraints." IEEE Transactions on Robotics 29.2 (2013): 346-362.
[3] Todorov, Emanuel, Tom Erez, and Yuval Tassa. "MuJoCo: A physics engine for model-based control." 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2012.