Discovering Optimal Natural Gaits of Dissipative Systems via Virtual Energy Injection
Korbinian Griesbauer, Davide Calzolari, Maximilian Raff, C. David Remy, Alin Albu-Schäffer · 2025
Read Discovering Optimal Natural Gaits of Dissipative Systems via Virtual Energy Injection by Korbinian Griesbauer, Davide Calzolari, Maximilian Raff, C. David Remy, Alin Albu-Schäffer (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-15513.txt.
Introduction
Korbinian Griesbauer, Davide Calzolari, Maximilian Raff, C. David Remy, and Alin Albu-Schäffer look for the natural gaits of legged robots that waste little energy. Legged machines handle rough ground well but usually use more energy than wheeled ones. Elastic parts can store and return energy, letting a robot ride its own natural motion instead of fighting it.
A new virtual energy injection method first finds passive motion patterns, even in systems that lose energy through friction or hard impacts. A continuation step then grows those patterns into energy-best controls for the fully powered robot. Tests in simulation cover single-leg and multi-leg systems and offer guidance for building more efficient elastic legged robots.
Catalog record
| CREATOR | Korbinian Griesbauer, Davide Calzolari, Maximilian Raff, C. David Remy, Alin Albu-Schäffer |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-15513
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Discovering Optimal Natural Gaits of Dissipative Systems via Virtual Energy Injection Korbinian Griesbauer, Davide Calzolari, Maximilian Raff, C. David Remy, Alin Albu-Schäffer Abstract Legged robots offer several advantages when navigating unstructured environments, but they often fall short of the efficiency achieved by wheeled robots. One promising strategy to improve their energy economy is to leverage their natural (unactuated) dynamics using elastic elements. This work explores that concept by designing energy-optimal control inputs through a unified, multi-stage framework. It starts with a novel energy injection technique to identify passive motion patterns by harnessing the system’s natural dynamics. This enables the discovery of passive solutions even in systems with energy dissipation caused by factors such as friction or plastic collisions. Building on these passive solutions, we then employ a continuation approach to derive energy-optimal control inputs for the fully actuated, dissipative robotic system. The method is tested on simulated models to demonstrate its applicability in both single- and multi-legged robotic systems. This analysis provides valuable insights into the design and operation of elastic legged robots, offering pathways to improve their efficiency and adaptability by exploiting the natural system dynamics. I Introduction The field of robotics is undergoing rapid advancements, with a particular emphasis on the development of humanoid robots that have gathered significant public attention. A key area of research focuses on developing robots capable of fast, safe, and efficient locomotion, modeled after human-like legs. A variety of possible motions enables such systems to move themselves forward. These periodic motions are called gaits . However, an essential concern is achieving economic movement. This goal is typically pursued through the use of optimal control [ 26 ] or more specifically through methods like Differential Dynamic Programming [ 14 ] . Virtual Energy Injection Solve Large NLP Root Finding Problem Homotopic Continuation Optimal Actuated Gait Actuated Fig. 1: The presented approach utilizes three main concepts. We first apply virtual energy injection to obtain a quasi-passive model. A quasi-passive system is not supported by actuation but through a non-physical energy input scaled by one scalar parameter. Through root-search , we find a quasi-passive gait. The homotopic continuation allows the derivation of a fully actuated optimal gait. Overall, this approach leverages the natural dynamics of the robotic system, thereby avoiding the need to solve large NLPs with high-dimensional decision variables. Still, several concepts can be investigated to increase the energy economy of the systems themselves and simplify the search process. One such concept is the application of elasticity within the system [ 28 , 27 ] . Similarly to tendons in animals’ legs, the springs can store elastic energy during the stance phase of running and release it during lift-off [ 1 ] . This allows the exploitation of nonlinear modes to execute highly efficient motions [ 6 , 8 ] . The elasticity is especially advantageous for legged systems, as the springs can absorb impacts to protect the motors. Furthermore, springs enable motions that are not supported by any actuation, which we call passive . Finding passive gaits is typically achieved by simplifying the system to exclude damping. The absence of damping enables the conservation of energy and the establishment of periodic motions, which are similar to the nonlinear modes of the mechanical system, without applying actuation [ 10 ] . An additional model simplification to sustain conservativeness is to take the limit of the foot mass going against zero (m_{\mathrm{F}}\rightarrow 0) , such that no energy is lost when the foot’s velocity is brought to zero at impact [ 11 ] . Conservative models also serve as ideal templates for generating energetically optimal gaits, as they enable locomotion with zero energy input [ 9 ] . However, the modeling process leading to energy conservation significantly influences the system dynamics, posing the question of how to adapt the gaits found for the conservative system to the real robot. In particular, the need for vanishing damping can introduce unwanted oscillations, while the requirement for zero collision mass may alter the system’s degrees of freedom [ 12 ] . The standard approach for exploiting the elasticity of a system remains optimal control [ 26 ] . However, there are several challenges associated with its implementation. The search space for a nonlinear program (NLP) increases drastically with the complexity of the system. This is due to the fact that the system states, as well as the actuation states, must be explored, resulting in a high-dimensional decision variable. Moreover, for the solver to produce reliable results, the initial guess provided by the user must be reasonably close to the actual solution. The identification of entire families of gaits [ 17 ] in advance may be a prerequisite in this case. This can be bypassed by a modular approach. In this paper, we seek to bridge between the passive template models and active optimal control. This will allow us to apply successful tools that exist for the analysis and generation of gaits in passive models [ 19 , 20 ] to a larger class of hybrid dynamic systems. The key concept that we introduce is quasi-passive models . In such models, an energy source that is parameterized with a single parameter \gamma counteracts damping, friction, and other losses. This energy source does not necessarily have to be physically realizable and should not be thought of as a controller. Its only purpose is to counteract energy dissipation to enable periodic locomotion. In a subsequent continuously-parameterized optimization process, this input is gradually reduced and replaced by an actuation profile, which is physically feasible. Through this approach, we are able to leverage the natural dynamics of the system for efficient gait generation. The concept of quasi-passive models includes passive dynamic walkers on a slope [ 16 , 22 ] . In these systems, the energy source is provided by the slope \gamma since it defines an external force (gravity) pulling the robot down the ramp. This work can be seen as a generalization of [ 22 ] , where families of gaits of zero actuation under the influence of gravity are found and connected to their optimally actuated counterparts. Similarly, Raff et al. [ 18 ] employ a simplified energetically conservative model to identify optimally actuated gaits through a model homotopy and numerical continuation on the optimality. The methods presented in [ 18 ] lay the foundation for our conceptual framework. Continuation is also useful to find entire families of gaits along solution manifolds, a process that is typically accomplished using energetically conservative models [ 17 , 4 ] . The novel energy input introduced in our work is motivated by the periodic motion concept in [ 13 ] , where the influence of damping onto nonlinear modes of a system is discussed. We contribute to the state of the art in efficient locomotion with elastic legged systems with the following: 1. A novel strategy to counteract energy losses in elastic locomotion using a virtual input, which is parameterized with a single parameter (Sec. III-B ). 2. A modular approach to find optimally actuated gaits for elastic legged models including damping and other losses. The framework consists of a root-search with direct collocation and a homotopic continuation on the first-order optimality (Sec. V