Loco-manipulation robot for surveillance in unstructured, hostile environment

Aim

Advancing loco-manipulation skills for robot surveillance in unstructured, dark, and hostile environments.

Objectives

  1. Enhancing loco-manipulation skills of legged robots.
  2. mproving and generalizing the versatility of loco-manipulation robots in unknown, contact-rich environments by potentially exploiting contact-implicit and adaptive approaches for planning, control, and estimation.
  3. Integrate results in open-source projects such as Crocoddyl and Pinocchio.
  4. Demonstrate the validity of our approach in multiple legged systems navigating under tunnels, hostile and highly constrained environments.

Description

Most autonomous sensing platforms are based on wheels. Even though they are versatile in urbanized environments, they lack adaptability in unstructured environments where human risk is high. Examples of these scenarios are damaged buildings or nuclear facilities. In these scenarios, legged robots can navigate effectively and exploit tactile sensing for mapping environments with a lack of visibility as in a mine.

This project focuses on advancing loco-manipulation navigation and mapping capabilities of legged robots. It will address the current gaps in whole-body motion control [1], [2] in which its hybrid nature doesn't scale well to contact-rich navigation and sensing. To enable this, we will develop contact-implicit approaches [3], [4], advance control and estimation solvers [5], [6] and simulators [6]. By combining these research efforts, this project aims at enabling versatile loco-manipulation navigation and autonomous sensing in challenging environments. Loco-manipulation navigation refers to multi-agent robots that leverage both arms and legs.

References

  1. Thomas Corberès, Carlos Mastalli, Wolfgang Merkt, Ioannis Havoutis, Maurice Fallon, Nicolas Mansard, Thomas Flayols, Sethu Vijayakumar, and Steve Tonneau. “Perceptive Locomotion through Whole-Body MPC and Optimal Region Selection”. Submitted to the International Journal of Robotics Research IJRR, 2023.
  2. Carlos Mastalli, Saroj Prasad Chhatoi, Thomas Corberès, Steve Tonneau, and Sethu Vijayakumar. "Inverse-Dynamics MPC via Nullspace Resolution. IEEE Transaction on Robotics (TRO), 2023.
  3. Patrick M Wensing and Micheal Posa and Yue Hu and Adriend Escande and Nicolas Mansard and Andrea Del Prete. "Optimization-based control for dynamic legged robots". IEEE Transaction on Robotics (TRO), 2023.
  4. Simon Le Cleac’h, Taylor Howell, Shuo Yang, Chi-Yee Lee, John Zang, Mac Schwager, and Zachary Manchester. "Fast Contact-Implicit Model Predictive Control". IEEE Transaction on Robotics (TRO), 2024.
  5. Carlos Mastalli, Rohan Budhiraja, Wolfgang Merkt, Guilhem Saurel, Bilal Hammoud, Maximilien Naveau, Justin Carpentier, Ludovic Righetti, Sethu Vijayakumar, Nicolas Mansard. “Crocoddyl: An efficient and versatile framework for multi-contact optimal control”. IEEE International Conference on Robotics and Automation (ICRA), 2020.
  6. Sergi Martinez, Robert Griffin, and Carlos Mastalli. “Multi-Contact Inertial Estimation and Localization in Legged Robots”. Submitted to. The IEEE Robotics and Automotion Letters (RAL), 2024
  7. Moritz Geilinger, David Hahn, Jonas Zehnder, Moritz Bächer, Bernhard Thomaszeski, and Stelian Coros. “ADD: Analytically Differentiable Dynamics for Multi-Body Systems with Frictional Contact”. ACM Transactions on Graphics (TOG), 2020.

Research theme: 

Principal supervisor: 

Dr Carlos Mastalli
Heriot-Watt University, School of Engineering & Physical Sciences
C.Mastalli@hw.ac.uk

Assistant supervisor: 

Dr Steve Tonneau
University of Edinburgh, School of Informatics
stonneau@ed.ac.uk