Thesis Work Robotics R&D Motion Control i Västerås

Join ABB and work in a team that is dedicated to creating a future where innovative digital technologies allow greater access to cleaner energy.

Within ABB Robotics R&D Motion Control department we are proposing several Master thesis next spring. The Motion Control department are responsible for a wide range of areas within the robot controller development spanning from modeling, identification, control design and as well as optimization for path planning and numerous other motion control functionalities.

When applying please specify which one the three Master thesis you are interested in:

1. Simulation model of a current controller
This thesis work consists of two parts. The first part is to implement a simulation model for a current controller and the drive unit hardware in a robot system. The model will be implemented in Simulink and it should be possible to use it for example for studying the system behavior when changing motors. The second part of this thesis work will be to develop an algorithm, this could for instance be field weakening or deadtime compensation. Validation of the model will be done by performing measurements on robot or motor cradle.

2. Comparative study on physics simulators for mobile manipulation
Simulation is an important step from design to system integration. This is even more critical in applications in which the task is solved by a policy learned with machine learning approaches, requiring high numbers of evaluations in a short time. The objective of the Thesis is to compare different simulators (open source and not), such as AGX (Algoryx), MuJoco, Gazebo, pyBullet and CoppeliaSim in terms of speed, parallelization capabilities, accuracy, ROS integration, model description etc. The first part of the work is to find and describe the available simulators, choosing the most promising ones that will be compared against the metrics in the second part. This requires designing a descriptor of the robot (e.g. the urdf for the Mobile YuMi) and the task to solve. An interesting research question is to see how the evaluated metrics compare to real learning on the systems, with some analysis of the trade-offs between different aspects such as speed v.s. fidelity. The results of this enquiry will support the choice of the right simulator for specific applications.

3. Development of a learning from demonstration system for mobile manipulators
In learning from demonstration, a human shows a robot how to solve a task, in different scenarios. Then the robot translates the perceived demonstration as a policy, aimed to solve the same task but in a slightly different (unseen) scenario. There are different representations of the policy and Behavior Tree is a promising one, having natural support for task hierarchy, reactivity, and modularity. Scope of the thesis is to first setup an existing system to learn from demonstration for a mobile YuMi, with the implementation of a state-of-the-art object recognition using an RGBD camera. The perceived demonstrations will be then encoded as leaves of a Behavior Tree. Finally, a task will be showed and the robot will learn a policy, represented as a Behavior Tree, solving the task.

Details:
• Period: January-June 2021
• Number of credits: Master thesis 30p
• Number of students: 1-2 for each Master thesis
• Location: Robotics R&D office at Finnslätten, Västerås

• The thesis work is independent, but you will be a part of an agile scrum team feeling the pulse of the daily work in the team.
• You will have a dedicated supervisor helping you in the right direction and a lot of possibilities to influence the way forward as you see fit.
• You will live ABB’s core values of safety and integrity, which means taking responsibility for your own actions while caring for your colleagues and the business.

• You are probably studying for a Master of Science, Engineering Physics, Electrical Engineering or Computer Science degree.
• Strong interest in Control theory or Machine learning.
• Sufficient knowledge of Matlab, C/C++, Python and real time embedded SW development.