Receding-Horizon Motion Planning of Quadrupedal Robot Locomotion

Aug 3, 2026

Upload by Ivy

Quadrupedal robots are designed to provide efficient, robust mobility on uneven terrain. This thesis develops a planning approach that unites numerical optimization with machine learning to obtain interpretable kinodynamic trajectories for agile natural gaits. The proposed method, RHECALL, solves nonlinear programs using learned initialization to generate terrain-aware whole-body motion with long-horizon fidelity on complex scenes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages205

Reading Time8 h 37 min

AccessFree to download

Introduction

Context and motivation

Aims and scope

Outline

Contributions

Publications

Background

Introduction

Overview of legged robots

Receding-Horizon Motion Planning of Quadrupedal Robot Locomotion Oliwier Aleksander Melon Mansfield College University of Oxford A thesis submitted for the degree of Doctor of Philosophy Trinity 2022 Declaration This thesis is submitted to the Department of Engineering Science, University of Oxford, in fulfilment of the requirements for the degree of Doctor of Philosophy. This thesis is entirely my own work, and except where otherwise stated, describes my own research. Oliwier Aleksander Melon, Mansfield College Acknowledgements I am deeply grateful to my co-supervisors, Prof. Maurice Fallon and Prof. Ioannis Havoutis, for giving me the unique opportunity, extensive resources and guidance to explore, learn and work on exciting research problems. To my examiners, Dr Nicolas Mansard and Prof. Paul Goulart, thank you foryour time, expertise and constructive feedback which helped me improve this thesis. I would also like to thank my TOS and COS assessors, Prof. Nick Hawes and Prof. Steve Roberts, for their insightful advice which helped shape this work. Heartfelt thanks to the ORI administrators, engineers, academics, researchers and students whom I had the pleasure of working with—especially to the DRS group and my fellow DPHIL student, Yiduo Wang, for sharing the ups and downs of this journey with me. I was extremely fortunate to have collaborated with Dr Mathieu Geisert, Dr David Surovik, Dr Romeo Orsolino and Dr Wolfgang Merkt. By virtue of their encouragement, support and contributions—including countless hours spent together discussing, planning, implementing, testing and documenting the research—I was able to reach this point. Mansfield College, its staff and students have been the source of wonderful memories. It was an honour to serve as MCR President in 2020–2021 and a pleasure working with many like-minded people over the years. Thank you tomy rowing friends at MCBC for all the great times on and off the river. I could not have completed these studies without my friends and family. I wish to thank my housemates for their company, especially Alexander Cliffe and Armands Strikis with whom I lived for a long time and share many good memories. I am so thankful to my partner, Ella Buys, for her unwavering love and invaluable support, and for being a world of difference during Covid and caring for me during the most testing periods. Special thanks to Alex and Ella for helping me review this thesis. Lastly, I am eternally grateful to my mother, Jowita Melon, for always doing everything she can to help me overcome any hurdles and achieve my goals. Funding: My gratitude also goes to UKRI EPSRC, EU Horizon 2020 projects THING and MEMMO, and Moog Inc. for the provision of financial support which made this research possible. Abstract Quadrupedal robots are designed to offer efficient and robust mobility on uneven terrain. This thesis investigates combining numerical optimization and machine learning methods to achieve interpretable kinodynamic planning of natural and agile locomotion. The proposed algorithm, called Receding-Horizon Experience-Controlled Adaptive Legged Locomotion (RHECALL), uses nonlinear programming (NLP) with learned initialization to produce long-horizon, high-fidelity, terrain-aware, whole-body trajectories. RHECALL has been implemented and validated on the ANYbotics ANYmal B and C quadrupeds on complex terrain. The proposed optimal control problem formulation uses the single-rigidbody dynamics (SRBD) model and adopts a direct collocation transcription method which enables the discovery of aperiodic contact sequences. To generate reliable trajectories, we propose fast-to-compute analytical costs that leverage the discretization and terrain-dependent kinematic constraints. To extend the formulation to receding-horizon planning, we propose a segmentation approach with asynchronous centre of mass (COM) and end-effector timings and a heuristic initialization scheme which reuses

FAQs

What problem does the thesis address for quadrupedal robots?

It targets interpretable kinodynamic planning that supports efficient, robust locomotion on uneven terrain, combining numerical optimization with machine learning.

What is RHECALL and what does it produce?

RHECALL stands for Receding-Horizon Experience-Controlled Adaptive Legged Locomotion. It uses nonlinear programming with learned initialization to generate long-horizon, high-fidelity, terrain-aware whole-body trajectories.

How does the work extend the approach to receding-horizon planning?

It proposes a segmentation method with asynchronous COM and end-effector timings, heuristic initialization that reuses the previous solution, and integration of real-time 2.5D perception for online foothold selection.

Aug 3, 2026

Upload by Ivy