Abstract
Balancing is a fundamental need for legged robots due to their unstable floating-base nature. Balance control has been thoroughly studied for simple models such as the linear inverted pendulum thanks to the concept of the instantaneous capture point (ICP), yet the constant center of mass height assumption limits the application. This paper explores balancing of the variable-height inverted pendulum (VHIP) model by introducing the instantaneous capture input (ICI), an extension of the ICP based on its key properties. Namely, the ICI can be computed as a function of the state, and when this function is used as the control policy, the ICI is rendered stationary and the system will eventually come to a stop. This characterization induces an analytical region of capturable states for the VHIP, which can be used to conceptually guide where to step. To further address state and control constraints during recovery, we present and theoretically analyze an explicit ICI-based controller with online optimal feedback gains. Simulations demonstrate the validity of our controller for capturability maintenance compared to an approach based on the divergent component of motion.
Introduction
Maintaining balance is a core challenge in legged robotics. Traditionally, the Linear Inverted Pendulum (LIP) model has served as the foundation for balance control, leading to the widely used Instantaneous Capture Point (ICP) concept. The ICP identifies a location on the ground where a robot must place its foot to come to a complete stop. However, the LIP model relies on the restrictive assumption that the robot's Center of Mass (CoM) remains at a constant height. In reality, robots often need to adjust their vertical position—crouching or extending—to recover from significant disturbances.
This paper addresses the limitations of constant-height models by focusing on the Variable-Height Inverted Pendulum (VHIP). While more realistic, the VHIP is inherently nonlinear and difficult to analyze. The authors propose the Instantaneous Capture Input (ICI), a theoretical extension of the ICP that provides a closed-form analytical characterization of "capturable" states for systems with varying height.
The Instantaneous Capture Input (ICI) Framework
The ICI is not merely a point on the ground but a control input pair consisting of the Zero Moment Point (ZMP) position ($p$) and the virtual leg stiffness ($\lambda$), which governs vertical force.
The core "magic" of the ICI lies in its stationarity. The authors derived a unique expression for the ICI based solely on the robot's current state (position and velocity). When this function is applied as the control policy, the ICI itself becomes stationary. This transformation effectively linearizes the deceleration process, allowing the robot to come to a stop predictably even within a nonlinear dynamical system. Unlike the Divergent Component of Motion (DCM), which requires tracking an auxiliary state, the ICI depends only on the instantaneous CoM state, making it highly suitable for real-time safety analysis.
Capturability Analysis and Control
A key contribution of this work is the analytical definition of the "capture basin"—the set of all states from which the robot can successfully recover. By leveraging the ICI, the authors established "inner" and "outer" approximations of this basin, creating a reliable safety map for controllers.
To manage real-world physical constraints, such as finite foot size and limited leg strength, the authors developed a feedback controller using sequential Linear Programming (LP). This controller optimizes feedback gains in real-time, ensuring the robot's Center of Pressure remains within its physical support polygon while maximizing its recovery potential.
Evaluation and Results
The proposed method was evaluated through high-fidelity simulations using parameters from the UBTECH Walker humanoid robot. In Monte Carlo trials involving 10,000 different push scenarios, the ICI-based controller demonstrated a significantly larger domain of attraction compared to state-of-the-art DCM-based approaches.
The ICI was particularly effective in scenarios involving high vertical velocities, where traditional models often fail. Furthermore, the analytical approximations of the capture basin were shown to be remarkably tight, providing an accurate boundary for determining robot safety during aggressive maneuvers.
Conclusion and Future Work
The Instantaneous Capture Input provides a powerful new tool for balancing robots that utilize vertical motion. By inducing linear CoM trajectories from nonlinear dynamics, it offers both theoretical elegance and practical robustness. Future research will focus on extending this 2D planar formulation to full 3D motion and incorporating centroidal dynamics to account for the angular momentum of robot limbs. Additionally, exploring multi-contact scenarios remains a promising avenue for enhancing humanoid stability in complex environments.
Summary updated