Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Closing the Sim-To-Real Gap for Automated Controller Tuning
The goal of this project is to develop a pipeline for automated controller tuning that leverages machine-learning based simulations of the real world.
Keywords: Planning & Control, Sim-To-Real, Automated Controller Tuning, Learning Control
Robust planning and control algorithms are critical for autonomous mobile robots to reliably navigate challenging environments. One key aspect for the optimal performance of these algorithms are well-tuned parameters that govern the effective navigation behavior of the robot. Tuning these parameters is a time-consuming process and has to be repeated for different robot types or under different environmental conditions. To avoid this overhead, we would ideally like to find the optimal parameters in simulated environments. However, due to the sim-to-real gap, the results from simulation do not necessarily translate to the real world.
The goal of this project is twofold: on the one hand, we would like to extend our simulation environment by a component that mimics the robot behavior based on a learned model using real-world data. To this end, you will investigate what type of machine learning model serves as a good trade-off between accuracy and computational requirements. After successful deployment of the learned model, the next goal is the development of a framework with which the parameters of our controller can be automatically tuned. The first part of the project therefore enables the tuning process to be performed in simulation so that the results translate to the real world.
You will be invited to contribute to a common framework and will have the chance to work with close-to-market robot prototypes. Finally, by the end of the project, you will have developed considerable knowledge in the very challenging topics of planning and control, supervised machine learning and black-box optimization algorithms. The hands-on experience with a mobile robotic platform will further augment this experience with a deep understanding of the software-hardware interplay.
What We Offer: Possibility to contribute to ongoing research in the exciting and quickly developing fields of planning & control and machine learning. Work with and be part of a team of enthusiastic roboticists and researchers in a Zurich based robotics startup in collaboration with the Autonomous Systems Lab, one of the largest robotic labs in the world. Possibility to deploy your algorithms to different robotic platforms and highly-valued hands-on experience.
Robust planning and control algorithms are critical for autonomous mobile robots to reliably navigate challenging environments. One key aspect for the optimal performance of these algorithms are well-tuned parameters that govern the effective navigation behavior of the robot. Tuning these parameters is a time-consuming process and has to be repeated for different robot types or under different environmental conditions. To avoid this overhead, we would ideally like to find the optimal parameters in simulated environments. However, due to the sim-to-real gap, the results from simulation do not necessarily translate to the real world.
The goal of this project is twofold: on the one hand, we would like to extend our simulation environment by a component that mimics the robot behavior based on a learned model using real-world data. To this end, you will investigate what type of machine learning model serves as a good trade-off between accuracy and computational requirements. After successful deployment of the learned model, the next goal is the development of a framework with which the parameters of our controller can be automatically tuned. The first part of the project therefore enables the tuning process to be performed in simulation so that the results translate to the real world.
You will be invited to contribute to a common framework and will have the chance to work with close-to-market robot prototypes. Finally, by the end of the project, you will have developed considerable knowledge in the very challenging topics of planning and control, supervised machine learning and black-box optimization algorithms. The hands-on experience with a mobile robotic platform will further augment this experience with a deep understanding of the software-hardware interplay.
What We Offer: Possibility to contribute to ongoing research in the exciting and quickly developing fields of planning & control and machine learning. Work with and be part of a team of enthusiastic roboticists and researchers in a Zurich based robotics startup in collaboration with the Autonomous Systems Lab, one of the largest robotic labs in the world. Possibility to deploy your algorithms to different robotic platforms and highly-valued hands-on experience.
- Make yourself familiar with our robotic planning & control and simulation frameworks as well as current state-of-the-art controller tuning solutions.
- Design, train and compare different machine learning models that emulate the robot’s behavior.
- Design, implement and test an automated controller tuning pipeline for simulation and hardware.
- Build upon the state of the art by developing your own ideas and your supervisor's input.
- Design and conduct experiments with a mobile robot to evaluate the selected approach.
- Make yourself familiar with our robotic planning & control and simulation frameworks as well as current state-of-the-art controller tuning solutions. - Design, train and compare different machine learning models that emulate the robot’s behavior. - Design, implement and test an automated controller tuning pipeline for simulation and hardware. - Build upon the state of the art by developing your own ideas and your supervisor's input. - Design and conduct experiments with a mobile robot to evaluate the selected approach.
- Strong self-motivation and curiosity for solving challenging robotic problems
- Previous experience in the fields of planning & control and machine learning
Excellent programming skills, ideally in Python and C++
- Experience with Linux, ROS, and typical development tools such as git are advantageous
- A very good academic record is desirable, but may be compensated by expert knowledge in the areas mentioned above
- Strong self-motivation and curiosity for solving challenging robotic problems - Previous experience in the fields of planning & control and machine learning Excellent programming skills, ideally in Python and C++ - Experience with Linux, ROS, and typical development tools such as git are advantageous - A very good academic record is desirable, but may be compensated by expert knowledge in the areas mentioned above
If you are interested, please send your transcripts and CV to Lukas Fröhlich (lukas.froehlich@sevensense.ch) and Kamil Ritz (kamil.ritz@sevensense.ch).
If you are interested, please send your transcripts and CV to Lukas Fröhlich (lukas.froehlich@sevensense.ch) and Kamil Ritz (kamil.ritz@sevensense.ch).