Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Dynamic Roadmap Planning for Nonholonomic Systems
In this project, we are hoping to speed up the performance of the sampling-based planners by combining them with the dynamic roadmaps.
Keywords: Planning, sampling, roadmap
Sampling based planning methods have emerged as very successful tools for planning, even in high dimensional spaces. In sampling-based algorithms, collision checking is usually the most expensive operation and reportedly consumes up to 90–95% of the planning time. Lazy collision checking is used to delay the collision checking until it is needed or limit it to particular regions. However, these techniques only reduce the collision checking time indirectly by reducing the number of calls rather than the actual computation time of the collision checking function. In contrast, the Dynamic Roadmap (DRM), algorithmically reduces the collision checking time by encoding configuration-to-workspace occupation information. Typically DRMs are used for planning in the area of manipulation. In this project, we would like to investigate their applicability to mobile robotics. We are interested in planning in the presence of nonholonomic constraints for mobile robotics(e.g. vehicle-like robots). The successful outcome of the project would result in faster planning times for problems in cluttered environments with mobile robots. If the time allows, the method could be extended for contact schedule planning with legged robots.
Literature:
- [1] Kallman, Marcelo, and Maja Mataric. "Motion planning using dynamic roadmaps." IEEE International Conference on Robotics and Automation, 2004. Proceedings. ICRA'04. 2004. Vol. 5. IEEE, 2004.
- [2] Yang, Yiming, et al. "HDRM: A resolution complete dynamic roadmap for real-time motion planning in complex scenes." IEEE Robotics and Automation Letters 3.1 (2017): 551-558.
Sampling based planning methods have emerged as very successful tools for planning, even in high dimensional spaces. In sampling-based algorithms, collision checking is usually the most expensive operation and reportedly consumes up to 90–95% of the planning time. Lazy collision checking is used to delay the collision checking until it is needed or limit it to particular regions. However, these techniques only reduce the collision checking time indirectly by reducing the number of calls rather than the actual computation time of the collision checking function. In contrast, the Dynamic Roadmap (DRM), algorithmically reduces the collision checking time by encoding configuration-to-workspace occupation information. Typically DRMs are used for planning in the area of manipulation. In this project, we would like to investigate their applicability to mobile robotics. We are interested in planning in the presence of nonholonomic constraints for mobile robotics(e.g. vehicle-like robots). The successful outcome of the project would result in faster planning times for problems in cluttered environments with mobile robots. If the time allows, the method could be extended for contact schedule planning with legged robots.
Literature:
- [1] Kallman, Marcelo, and Maja Mataric. "Motion planning using dynamic roadmaps." IEEE International Conference on Robotics and Automation, 2004. Proceedings. ICRA'04. 2004. Vol. 5. IEEE, 2004.
- [2] Yang, Yiming, et al. "HDRM: A resolution complete dynamic roadmap for real-time motion planning in complex scenes." IEEE Robotics and Automation Letters 3.1 (2017): 551-558.
- Literature research on sampling based planning, dynamic roadmaps, motion primitives, Reed Shepp curves
- Choosing the right data structure for repeated lookups in 2D space.
- Implementing the planner
- Setting up the benchmark environment
- Benchmarking against existing planning algorithms
- Literature research on sampling based planning, dynamic roadmaps, motion primitives, Reed Shepp curves - Choosing the right data structure for repeated lookups in 2D space. - Implementing the planner - Setting up the benchmark environment - Benchmarking against existing planning algorithms
- Excellent programming skills (C++ prefered)
- Knowledge of planning algorithms
- Basic knowledge of data structures
- Knowledge of ROS is a plus
- Excellent programming skills (C++ prefered) - Knowledge of planning algorithms - Basic knowledge of data structures - Knowledge of ROS is a plus
Edo Jelavic, jelavice@ethz.ch
Johannes Pankert, johannes.pankert@mavt.ethz.ch
Please send your grade transcripts and CV.