Design and implementation of a minimalistic setup for manipulation research

Recent advances in numerical optimization made it feasible to cast the problem of autonomous robotic manipulation as an optimal control problem [1]. Here, manipulation goals are specified on a high level – e.g., as a target object pose or trajectory – and the specifics of the movement are left to the optimizer who generates corresponding control commands. This paradigm requires accurate modeling of the robot and the environment and therefore has mainly been verified in simulation so far.

The goal of this thesis is to design and assemble a testbed for manipulation research in the context of the problem outlined above. The envisioned system will consist of a 2 DoF planar manipulator and a gripper. The focus of the design is to allow accurate physics modeling, i.e, to limit sources of uncertainty (such as friction) as much as possible. An optional extension of the thesis is to incorporate a real-time perception system for target object pose detection in 2D.

Tasks:

• Design, assembly and modeling of a 2D manipulator
• Joint-level effort, velocity and position control
• Optionally, incorporate a real-time target object perception system

Necessary skills:

• Experience in mechanical design
• Experience in mechatronics
• Knowledge in multi-body systems
• Low-level control basics
• Experience in the ROS ecosystem and C++ programming would be helpful

 

Contact: Robert Krug

[1]  Y. Tassa, T. Erez and E. Todorov, "Synthesis and stabilization of complex behaviors through online trajectory optimization," 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2012, pp. 4906-4913.