Differentiable Invariant Sets for Hybrid Limit Cycles with Application to Legged Robots
Varun Madabushi, Akash Harapanahalli, Samuel Coogan, Maegan Tucker · 2026
Read Differentiable Invariant Sets for Hybrid Limit Cycles with Application to Legged Robots by Varun Madabushi, Akash Harapanahalli, Samuel Coogan, Maegan Tucker (2026) 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-2604-05108.txt.
Introduction
Madabushi and three co-authors develop differentiable invariant sets for hybrid limit cycles, for legged robots. For hybrid systems with behavior, the invariant set containing the limit cycle offers a way to study robustness. Computing these sets is expensive, especially for systems such as legged robots. The authors extend methods for overapproximating reachable sets using embeddings. They compute a set around the trajectory of a bipedal robot model. Their approach computes a reachable set around the flow, catalogs guard-surface intersections, and passes them through the reset map.
Catalog record
| CREATOR | Varun Madabushi, Akash Harapanahalli, Samuel Coogan, Maegan Tucker |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45539 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.05108
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Differentiable Invariant Sets for Hybrid Limit Cycles with Application to Legged Robots Varun Madabushi, Akash Harapanahalli, Samuel Coogan, Maegan Tucker Abstract For hybrid systems exhibiting periodic behavior, analyzing the invariant set containing the limit cycle is a natural way to study the robustness of the closed-loop system. However, computing these sets can be computationally expensive, especially when applied to contact-rich cyber-physical systems such as legged robots. In this work, we extend existing methods for overapproximating reachable sets of continuous systems using parametric embeddings to compute a forward-invariant set around the nominal trajectory of a simplified model of a bipedal robot. Our three-step approach (i) computes an overapproximating reachable set around the nominal continuous flow, (ii) catalogs intersections with the guard surface, and (iii) passes these intersections through the reset map. If the overapproximated reachable set after one step is a strict subset of the initial set, we formally verify a forward invariant set for this hybrid periodic orbit. We verify this condition on the bipedal walker model numerically using immrax , a JAX-based library for parametric reachable set computation, and use it within a bi-level optimization framework to design a tracking controller that maximizes the size of the invariant set. I Introduction Forward invariance plays a central role in analyzing the robustness of nonlinear dynamical systems [ 9 , 23 ] . However, for hybrid systems (i.e., those that exhibit both continuous dynamics and discrete transitions), accurately characterizing invariant sets is particularly challenging due to the presence of discontinuities and switching behaviors. These systems arise naturally in applications such as legged locomotion [ 25 ] , robotic manipulation with intermittent contact [ 26 ] , and mechanical systems with impacts or frictional constraints [ 8 ] . The study of invariant sets and regions of attraction (RoAs) have been approached through a wide range of techniques, including Lyapunov-based methods [ 15 , 23 ] , sum-of-squares (SOS) programming [ 24 , 18 , 6 ] , and Hamilton–Jacobi (HJ) formulations [ 19 ] . However, the practical applications have remained limited due to the computational complexity of the existing approaches and thus their limitation towards real-world high-dimensional systems. In 2011, Manchester [ 18 ] presented one of the first algorithmic approaches for computing inner estimates of the regions of attraction of limit cycles of a nonlinear hybrid system using iterative sum-of-squares (SoS) programming. Their approach formulated Lyapunov inequalities as semidefinite programs and was later demonstrated on the compass-gait walker, a low-dimensional (2-DOF, four states) nonlinear hybrid system [ 17 ] . Despite its theoretical elegance, SoS programming is known to scale poorly with system dimension, as the polynomial degree and number of decision variables grow combinatorially. Consequently, SoS-based methods are limited to relatively low-dimensional systems, typically below ten state variables, restricting their practical applicability to complex models. In 2022, Choi et. al. [ 7 ] instead cast the RoA computation problem as one of backward reachable set computation within the Hamilton–Jacobi (HJ) reachability framework. Their work generalized HJ formulations to accommodate hybrid dynamics, successfully recovering larger and more accurate RoA estimates compared to SoS-based approaches. However, the curse of dimensionality inherent in HJ partial differential equations remains a major bottleneck, as the computational complexity grows exponentially with the dimension of the state space. To address this challenge of scalability, our work instead quickly computes a parametric overapproximation of the reachable set using a set propagation approach. Approaches in this category [ 1 ] include generator-based set representations like zonotopes [ 2 ] , or Taylor models bounding the dynamics and/or flow [ 3 ] , and promise scalability to high-dimensional dynamical systems [ 13 ] . In this work, we use linear inclusions to propagate ellipsoidal reachable sets. This is achieved through immrax [ 12 ] , an efficient implementation of interval analysis [ 14 ] that exploits the hardware-accelerated and automatic differentiation capabilities of JAX [ 4 ] to enable scalable set propagation. However, this approach is currently limited to continuous-time systems. Extending this framework to hybrid systems with discrete impact events addresses new challenges: 1) accurately detecting and enclosing impact times within a continuous set evolution, 2) propagating sets across discrete reset maps that can be highly nonlinear, and 3) maintaining tight enclosures to avoid exponential growth through mode transitions. In this work, we build upon the immrax toolbox to address these challenges. The contributions of this work include the following: 1. We extend the theory of parametric reachability to identify conditions under which the parametric reachable set of a hybrid system is forward invariant. 2. We introduce a procedure using interval analysis to efficiently verify the forward invariance of a given set. 3. We demonstrate this methodology on a simplified planar walker model and showcase how the differentiability of immrax can be leveraged for control design in a bi-level optimization problem. II Preliminaries II-A Hybrid System Model In this work, we will consider an autonomous hybrid system with one continuous domain and one discrete impact event. It has been shown in prior work that multi-domain hybrid systems can be treated as an extension of this simpler representation [ 22 ] . Explicitly, the open-loop hybrid control system is represented as: \displaystyle\dot{x} \displaystyle=f(x,u), \displaystyle\quad x^{-} \displaystyle\in\mathcal{X}\setminus\mathcal{S}, (1a) \displaystyle x^{+} \displaystyle=\Delta(x^{-},v) \displaystyle\quad x^{-} \displaystyle\in\mathcal{S}, (1b) where x\in\mathcal{X}\subseteq\mathbb{R}^{n} is the system state, u\in\mathcal{U}\subseteq\mathbb{R}^{p} is a control input, f:\mathcal{X}\times\mathcal{U}\to\mathbb{R}^{n} defines a parameterized C^{1} vector field describing the continuous flow until it reaches the guard surface (also known as the switching surface) \mathcal{S}\subseteq\mathcal{X} , at which point the system undergoes an instantaneous jump from x^{-} to x^{+} via the reset map \Delta:\mathcal{X}\times\mathcal{V}\to\mathcal{X} under the discrete input v\in\mathcal{V}\subseteq\mathbb{R}^{q} . Under the control laws u=\pi^{c}(x) and v=\pi^{d}(x^{-}) , the closed-loop hybrid control system is represented as: \displaystyle\dot{x} \displaystyle=f_{\mathrm{cl}}(x):=f(x,\pi^{c}(x)), \displaystyle\quad x^{-} \displaystyle\in\mathcal{X}\setminus\mathcal{S} (2a) \displaystyle x^{+} \displaystyle=\Delta_{\mathrm{cl}}(x^{-}):=\Delta(x^{-},\pi^{d}(x^{-})) \displaystyle\quad x^{-} \displaystyle\in\mathcal{S} (2b) A solution x(t) of the closed-loop hybrid system is a right-continuous function satisfying ( 2 ), where x^{-}(t)=\lim_{\tau\nearrow t}x(\tau) and x^{+}(t)=\lim_{\tau\searrow t}x(\tau) . We assume that the guard surface is given as follows: for some C^{1} smooth h:\mathcal{X}\to\mathbb{R} , \displaystyle\mathcal{S}=\{x\in\mathcal{X}:h(x)=0,\ \dot{h}(x)<0\}, (3) where \dot{h}(x)=\mathcal{L}_{f_{cl}}h(x) . We also notationally set \{h\sim 0\}=\{x\in\mathcal{X}:h(x)\sim 0\} for the relations \sim\ \in\{=,\leq,\geq,<,>\} . Under mild conditions, the closed-loop hybrid system ( 2 ) has a unique right-continuous trajectory, which we assume does not have Zeno phenomenon. A periodic orbit of the closed-loop hybrid system is a solution x(t) of ( 2 ) that is periodic for some T , i.e., x(t)=x(t+T) . This periodicity condition can also be written using flow notation as: \displaystyle\Delta(\varphi_{T_{I}(x^{*})}(x^{*}))=x^{*}, (4) where x^{*}\in\Delta(\mathcal{S}) is a fixed point lying on the image of the guard surface, \varphi_{t}(x_{0}) denotes in flow notation the solution of our closed-loop continuous dynamics ( 2a ) at time t\in\mathbb{R}_{\geq 0} given the initial condition x_{0} , and T:=T_{I}(x^{*}) for the time-to-impact function T_{I}:\mathcal{X}\to\mathbb{R}\cup\{\infty\} , \displaystyle T_{I}(x)=\inf\{t\geq 0\mid\varphi_{t}(x)\in\mathcal{S}\}, (5) where we use the convention \inf\emptyset=\infty . The flow notation can also be used to denote an associated periodic orbit: \displaystyle\mathcal{O}:=\{\varphi_{t}(x^{*})\in\mathcal{X}\mid 0\leq t\leq T_{I}(x^{*})=T\} (6) where ( 6 ) is periodic if x^{*} satisfies ( 4 ). Stability of this periodic orbit is typically analyzed using the Poincaré return map [ 21 ] . In this work, we will instead use a generalization of