Learning Representations for Planning and Control

November 8, 2019 at IROS, Macau


  • Room Location: LG-R8
  • Instructions for accepted papers presentations have been updated. All accepted papers will have oral presentations!
  • Accepted papers are now available on our website!
  • Tentative workshop plan has been posted!


All deadlines are in Pacific Standard Times (PST)

Extended Abstract Submission deadline: August 1, 2019, 6 pm August 20, 2019, 6 pm

  • Acceptance Notification: September 20, 2019, 6 pm
  • Camera-ready submission deadline: October 5, 2019, 6 pm


Planning algorithms for control, also known as Motion Planning, has a long history ranging from methods with complete to probabilistic worst-case guarantees. However, despite having deep roots in artificial intelligence, these methods tend to be computationally inefficient in high-dimensional problems. On the other hand, machine learning advancements have led toward the systems that can perform complex decision-making by directly using the raw sensory information, thanks advancement in function approximation. This workshop aims to bring these two long-lived research communities under one forum to share insights towards building computationally tractable planning methods while retaining the theoretical guarantees.

Motion planning is the amalgamation of several vital components working together to generate a feasible plan which includes, but not limited to, sample generation, collision detection, nearest neighbor search, best edge selection, pruning, and rewiring. In this workshop, we aim to formalize the challenges and advantages of merging the machine learning methods with motion planning. We also seek to address how the advanced machine learning techniques can be leveraged to provide data-driven solutions to bottlenecks of motion planning. In general, the workshop will revolve around the following themes.

    • Identify and formalize bottlenecks of existing motion planning techniques and their solution through machine learning.
    • Highlight challenges of combining the two fields and arrive at potential research directions.
    • Address critical problems that could emerge from merging two fields and think about possible solutions.

Topics of Interest

A list of topics addressed in the workshop:

    • Data-driven approaches to motion planning.
    • Learning-based adaptive sampling methods.
    • Learning models for planning and control.
    • Imitation learning for planning and control.
    • Learning generalizable and transferable planning models.
    • Representation learning for planning.
    • Learning-based collision detection, edge selection and pruning techniques, and related topics.
    • Data-efficiency in data-driven techniques to planning
    • Formal guarantees to machine learning based planning methods.
    • Learning methods for Hierarchical planning such task and motion planning, and related topics.
    • Active learning methods for planning and related topics.

Paper Submission

We invite extended abstracts (minimum 2 pages) followed by camera-ready submission of accepted papers (no longer than 8 pages including references). All papers should follow the IEEE Conference Templates [Latex, MS Word]. Submissions can be original research, late-breaking results, or a literature review that fall under the scope of the workshop. All submissions should be made through the following link: https://cmt3.research.microsoft.com/LRPC2019.

Review Process:

All papers will be reviewed via single-blind review process: authors declare their names and affiliations in the manuscript for the reviewers to see, but reviewers do not know each other's identities, nor do the authors receive information about who has reviewed their manuscript. The papers acceptance decision will be based on contribution, novelty, and overall content.

Paper presentation:

All accepted papers will have oral presentations. Each speaker will have 8 minutes for their talk and 2 mins for the Q&A session. The changeover between speakers will happen during the Q&A session. Please bring your laptop for the presentation. We are expecting to have a projector (16:9) with a VGA connector.

Invited Speakers' Presentation Slides

Tentative Plan

Invited Speakers

Waseda University

University of North Carolina

UC San Diego

TU Darmstadt

NVIDIA Research

University of Hong Kong

University of Washington

List of Accepted Papers

  1. PaperID 3: D. Nass, B. Belousov, and J. Peters , "Entropic Risk Measure in Policy Search."
  2. PaperID 5: C. V. Perico, W. Decre , J. D. Schutter, and E. Aertbelien, "Learning basis functions from human demonstrations to plan generic reactive behaviors for different robot manipulation applications ."
  3. PaperID 6: Y. Deng, F. Hu, X. Zhang, X. Wu, and D. Luo, "Online Motion Generation for Bipedal Robot Stair Climbing."
  4. PaperID 8: Y. Fu , C. Theurer , J. Reimann, and S. Sen, "Learning Low-Dimensional Spatiotemporal Representations from Raw Video Inputs with LSTM-LSTM."
  5. PaperID 11: O. Celik, H. Abdulsamad, and J. Peters, "Chance-Constrained Trajectory Optimization for Non-linear Systems with Unknown Stochastic Dynamics."
  6. PaperID 12: M. Lutter, and J. Peters, "Lagrangian Mechanics and Conservation of Energy as Inductive Bias for Model Learning."
  7. PaperID 13: I. Akinola and P. Allen, "End-to-End Learning-Based Hierarchical Path Planning."
  8. PaperID 14: R. Reinhart, T. Dang, E. Hand, and K. Alexis, "Learning-based Path Planning for Autonomous Exploration of Subterranean Environments."
  9. PaperID 15: E. Camci and E. Kayacan, "End-to-End Motion Planning of Quadrotors Using Deep Reinforcement Learning."
  10. PaperID 17: D. McConachie and D. Berenson, "Learning When Not To Trust Your Model For Deformable Object Manipulation Planning."
  11. PaperID 19: S. Chinchali, A. Anemogiannis, and M. Pavone, "Task-Specific Representations for Robotic Perception and Control: Applications to Cloud Robotics."
  12. PaperID 20: D. Son, H. Yang and D. Lee, "Learnable Environment Model with Data Efficiency for MPC of Assembly Tasks."


Ahmed Qureshi

UC San Diego

Michael Yip

UC San Diego

Jan Peters

TU Darmstadt

Byron Boots

Georgia Tech

Dmitry Berenson

University of Michigan

Dorsa Sadigh

Stanford University

Marco Pavone

Stanford University


Ahmed Qureshi