Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Sampling-based Planning with Riemannian Motion Policies
Riemannian Motion Policies (RMPs) are a recently proposed modular, reactive planning framework. In this work, we want to investigate methods to synthesize longer horizon motion plans from an RMP.
Keywords: robotics, planning
The goal of this project is to investigate methods for efficiently synthesizing motion plans from Riemannian Motion Policies (RMPs) \[1]. RMPs are a recently proposed modular and reactive motion planning framework. The modular architecture enables the design of complex robot behaviors by combining multiple simpler policies in a mathematically rigorous way. Examples of atomic RMP policies are goal-reaching, obstacle avoidance, ensuring kinematic constraints [1], or surface following [2]. Unfortunately, the local and reactive nature of RMPs makes them prone to getting trapped in local minima. In this project, we want to leverage the strengths of RMPs for generating long-term motion plans, for example by querying it from a sampling-based planner such as RRT* [3] or MPPI [4]. A possible application could be 6DoF planning for our OMAVs in confined spaces.
1. Riemannian Motion Policies, N.D. Ratliff et al., 2018
2. Mesh Manifold Based Riemannian Motion Planning for Omnidirectional Micro Aerial Vehicles, M. Pantic et al., 2021
3. Sampling-based Algorithms for Optimal Motion Planning. Karaman, Sertac & Frazzoli, Emilio., 2011
4. Information-theoretic MPC for model-based reinforcement learning, G. Williams et al., 2017
The goal of this project is to investigate methods for efficiently synthesizing motion plans from Riemannian Motion Policies (RMPs) \[1]. RMPs are a recently proposed modular and reactive motion planning framework. The modular architecture enables the design of complex robot behaviors by combining multiple simpler policies in a mathematically rigorous way. Examples of atomic RMP policies are goal-reaching, obstacle avoidance, ensuring kinematic constraints [1], or surface following [2]. Unfortunately, the local and reactive nature of RMPs makes them prone to getting trapped in local minima. In this project, we want to leverage the strengths of RMPs for generating long-term motion plans, for example by querying it from a sampling-based planner such as RRT* [3] or MPPI [4]. A possible application could be 6DoF planning for our OMAVs in confined spaces.
1. Riemannian Motion Policies, N.D. Ratliff et al., 2018 2. Mesh Manifold Based Riemannian Motion Planning for Omnidirectional Micro Aerial Vehicles, M. Pantic et al., 2021 3. Sampling-based Algorithms for Optimal Motion Planning. Karaman, Sertac & Frazzoli, Emilio., 2011 4. Information-theoretic MPC for model-based reinforcement learning, G. Williams et al., 2017
1. Literature review on reactive planning and sampling-based planning with kinematic constraints
2. Build a prototype sampling-based RMP planner
3. Evaluate performance empirically
1. Literature review on reactive planning and sampling-based planning with kinematic constraints 2. Build a prototype sampling-based RMP planner 3. Evaluate performance empirically
- Highly motivated and independently working student
- Programming experience in C++ and Python
- Experience in robotic motion planning would be beneficial
- Strong mathematical background in analysis and probability would be beneficial
- Highly motivated and independently working student - Programming experience in C++ and Python - Experience in robotic motion planning would be beneficial - Strong mathematical background in analysis and probability would be beneficial
- Michael Pantic michael.pantic@mavt.ethz.ch
- Nikhilesh Alatur nikhilesh.alatur@mavt.ethz.ch
- Dr. Olov Andersson olov.andersson@mavt.ethz.ch
- Michael Pantic michael.pantic@mavt.ethz.ch - Nikhilesh Alatur nikhilesh.alatur@mavt.ethz.ch - Dr. Olov Andersson olov.andersson@mavt.ethz.ch
Sampling-Based-RMP.pdf
