Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Internship or Master Thesis: Shared-Control for Autonomous Wheelchair Robot Navigation
In this project, we will develop an autonomous wheelchair control for a novel wheelchair with omnidirectional control and test it in a final product. You would develop an algorithm for enabling shared-control navigation for omnidirectional control under complex environments using the architecture proposed in [2] and perform an evaluation with users at a large airport.
Autonomous navigation in highly populated areas remains a challenging task for robots because of the difficulty in guaranteeing safe interactions with pedestrians in unstructured situations. Application in wheelchair navigation is especially important for removing the cognitive load on users from the small task of avoiding obstacles, allowing them to focus on higher-level planning.
Improvements in deep learning and computer vision have allowed an increase in speed in the prediction of human position. This allows almost real-time image evaluation of the environment, and combining different sensors improves real-time behaviour in experiments. Many classical navigation algorithms still struggle to handle the prediction of human intention. The recent algorithm started combining crowd models with local robot sensor data [3] with success in implementation for crowd navigations [1,2].
[1] Paez-Granados, D., He, Y., Gonon, D., Jia, D., Leibe, B., Suzuki, K., & Billard, A. (2022). Pedestrian-Robot Interaction on Crowd Navigation: Reactive Control Methods and Evaluation. IEEE Conference International on Intelligent Robots and Systems (IROS), 1–8. https://github.com/epfl-lasa/crowdbot-evaluation-tools
[2] Paez-Granados, D., Gupta, V., & Billard, A. (2022). Unfreezing Social Navigation : Dynamical Systems based Compliance for Contact Control in Robot Navigation. IEEE International Conference on Robotics and Automation (ICRA), 1(1), 8368–8374. https://doi.org/10.1109/ICRA46639.2022.9811772
[3] Gonon, D. J., Paez-Granados, D., & Billard, A. (2021). Reactive Navigation in Crowds for Non-holonomic Robots with Convex Bounding Shape. IEEE Robotics and Automation Letters, 6(3), 4728–4735. https://doi.org/10.1109/LRA.2021.3068660
Autonomous navigation in highly populated areas remains a challenging task for robots because of the difficulty in guaranteeing safe interactions with pedestrians in unstructured situations. Application in wheelchair navigation is especially important for removing the cognitive load on users from the small task of avoiding obstacles, allowing them to focus on higher-level planning. Improvements in deep learning and computer vision have allowed an increase in speed in the prediction of human position. This allows almost real-time image evaluation of the environment, and combining different sensors improves real-time behaviour in experiments. Many classical navigation algorithms still struggle to handle the prediction of human intention. The recent algorithm started combining crowd models with local robot sensor data [3] with success in implementation for crowd navigations [1,2].
[1] Paez-Granados, D., He, Y., Gonon, D., Jia, D., Leibe, B., Suzuki, K., & Billard, A. (2022). Pedestrian-Robot Interaction on Crowd Navigation: Reactive Control Methods and Evaluation. IEEE Conference International on Intelligent Robots and Systems (IROS), 1–8. https://github.com/epfl-lasa/crowdbot-evaluation-tools
[2] Paez-Granados, D., Gupta, V., & Billard, A. (2022). Unfreezing Social Navigation : Dynamical Systems based Compliance for Contact Control in Robot Navigation. IEEE International Conference on Robotics and Automation (ICRA), 1(1), 8368–8374. https://doi.org/10.1109/ICRA46639.2022.9811772
[3] Gonon, D. J., Paez-Granados, D., & Billard, A. (2021). Reactive Navigation in Crowds for Non-holonomic Robots with Convex Bounding Shape. IEEE Robotics and Automation Letters, 6(3), 4728–4735. https://doi.org/10.1109/LRA.2021.3068660
- Using state-of-the-art algorithms to detect humans from sensor data. (Deep learning from images/lidar).
- Implement the sensing pipeline with integration to ROS as proposed in [2]
- Creating the interface of communication between low-high level controllers and ensuring a safe-proof system.
- Develop an interface of control for omnidirectional shared control that applies a dynamical system for guiding the motion.
- Performing live tests and adapting the code/hardware based on results.
- Developing a well-documented code and easily editable system through a modular architecture.
- Using state-of-the-art algorithms to detect humans from sensor data. (Deep learning from images/lidar). - Implement the sensing pipeline with integration to ROS as proposed in [2] - Creating the interface of communication between low-high level controllers and ensuring a safe-proof system. - Develop an interface of control for omnidirectional shared control that applies a dynamical system for guiding the motion. - Performing live tests and adapting the code/hardware based on results. - Developing a well-documented code and easily editable system through a modular architecture.
• Hands-on experience in both hardware and software for developing a novel system.
• Access to expertise from both control and navigation systems with direct application to industrial usage.
• Working with the potential of long-term applications from the results of this thesis work.
• Hands-on experience in both hardware and software for developing a novel system. • Access to expertise from both control and navigation systems with direct application to industrial usage. • Working with the potential of long-term applications from the results of this thesis work.
- Student of a Swiss university or university of applied sciences (ETH Zurich, EPFL, etc.) - Students at EU universities might be considered
- ETHZ: D-MAVT, D-INFK / EPFL: IMT, CS (or equivalent)
- Understanding of Embedded computing (preferable)
- Experience in ROS
- Knowledge of localization and mapping algorithms
- Knowledge of virtual environments (conda / docker)
- Strong experience with Python / C++
- Structured and reliable working style
- Ability to work independently on a challenging topic
- Student of a Swiss university or university of applied sciences (ETH Zurich, EPFL, etc.) - Students at EU universities might be considered - ETHZ: D-MAVT, D-INFK / EPFL: IMT, CS (or equivalent) - Understanding of Embedded computing (preferable) - Experience in ROS - Knowledge of localization and mapping algorithms - Knowledge of virtual environments (conda / docker) - Strong experience with Python / C++ - Structured and reliable working style - Ability to work independently on a challenging topic
Please send your CV, example code or project and the latest transcript of records to Dr. Diego Paez (diego.paez@hest.ethz.ch)
Please send your CV, example code or project and the latest transcript of records to Dr. Diego Paez (diego.paez@hest.ethz.ch)