Pierre-Alexandre Leziart

Pierre-Alexandre Leziart

PostDoc in Robotics

CNRS-AIST Joint Robotics Laboratory

About me

Postdoctoral fellow focusing on the design, implementation and improvement of state-of-the-art control architectures for legged robots. My research activities are motivated by all the challenges that remain to be tackled to control such complex mechatronic systems in a reliable way in real life situations. The advent of machine learning techniques in recent years has brought plenty of promising results, to the level of what has been achieved by more classical model-based approaches. I am convinced that performances can be pushed even further by combining both approaches to benefit from the best of both worlds.

Interests
  • Legged Robotics
  • Model Predictive Control
  • Reinforcement Learning
  • Torque Control
  • Optimal Control
Education
  • PhD in Robotics, 2022

    LAAS-CNRS and University of Toulouse, France

  • MEng in Mechatronics, 2019

    École Normale Supérieure de Rennes, France

  • Agrégation of Industrial Engineering Sciences, Electrical specialization, 2018

    French Ministry of Higher Education and Research

Recent Publications

Quickly discover relevant content by filtering publications.
(2024). CaT: Constraints as Terminations for Legged Locomotion Reinforcement Learning. ArXiv.

PDF Cite Code Video Website

(2023). A Hierarchical Scheme for Adapting Learned Quadruped Locomotion. IEEE-RAS 22nd International Conference on Humanoid Robots (Humanoids).

PDF Cite Video

(2023). Controlling the Solo12 quadruped robot with deep reinforcement learning. Scientific Reports.

Cite

(2022). CosySlam: investigating object-level SLAM for detecting locomotion surfaces.

Cite URL

(2022). Improved Control Scheme for the Solo Quadruped and Experimental Comparison of Model Predictive Controllers. IEEE Robotics and Automation Letters (RA-L).

Cite

Experience

 
 
 
 
 
Postdoctoral Fellow
May 2024 – Present Tsukuba, Japan
My PostDoc has been been focused on exploring legged locomotion for humanoid robots using machine learning approaches. This work is still in its preliminary phase.
 
 
 
 
 
Postdoctoral Fellow
November 2023 – April 2024 Toulouse, France
My first PostDoc focused on getting the Sassa quadruped robot ready for the IEEE ICRA 2023 Quadruped Robot Challenge using a reinforcement learning scheme. After that, I kept exploring reinforcement learning both for Sassa and Solo-12 to get more experience in this rising field for legged robot locomotion.
 
 
 
 
 
PhD Student
September 2019 – October 2023 Toulouse, France
My PhD focused on model-based walking architecture for the quadruped robot Solo-12 and centered on a centroidal model predictive controller followed by an instantaneous whole-body inverse kinematics. After the implementation of a whole control architecture based on the state of the art, I proposed and tested possible improvements. I also collaborated with several team members in their respective research, including reinforcement learning, 3D footstep planning, simultaneous localisation and mapping.
 
 
 
 
 
Master Thesis
February 2019 – July 2019 Lausanne, Switzerland
This master thesis involved the development of an algorithm to detect and avoid obstacles with a Ridgeback robotic platform. I had to implement the whole control scheme from low-level data retrieval with sensors to high-level data processing to build a map of the environment, which was successfully deployed on real hardware using ROS with real time performances.
 
 
 
 
 
Master Project
October 2018 – January 2019 Lausanne, Switzerland
This semester project was focused on the analysis of data from inertia measurement units that had been worn by meerkats in order to detect behavioural patterns such as feeding, foraging, grooming or looking out for predators. I used data-processing tool to spot and classify these activities in an automated way to remove the need for tedious hand-labeling.
 
 
 
 
 
Internship
May 2017 – August 2017 Lausanne, Switzerland
During this summer internship I was tasked with developing a set of experiments to highlight the potential of Roombots robots. I demonstrated their capabilities to build adaptive and self-organizing furniture as well as how they could assist elderly people. These experiments later served as proofs of concept for the Roombot project report.
 
 
 
 
 
Internship
May 2016 – July 2016 Rennes, France
This internship was focused on the use of a passive haptic device to provide contact feedback to users in virtual environments. I designed several use cases to test its limits and assess its performances.

Education & Qualifications

PhD in Robotics
  • Locomotion control of a lightweight quadruped robot
  • Under the supervision of Philippe Souères and Thomas Flayols
  • Issued by the University Toulouse 3 Paul Sabatier
  • Defended October 17, 2022.
Master of Engineering
  • Engineering of Complex Systems
  • Academic exchange at Ecole Polytechnique Fédérale de Lausanne, Switzerland.
Agrégation
  • Industrial Engineering Sciences, electrical specialization
  • External exam, rank: 1st
Master of Engineering
  • Engineering of Complex Systems

Contact

Feel free to contact me if you have any question about my work!