Introduction
Bipedal robot locomotion relies heavily on the foot-ground interface. While traditional designs like point feet or flat feet offer simplicity or static stability, they often fail to capture the dynamic efficiency of human walking. Human feet utilize a complex, segmented rollover shape—comprising the hind, mid, and fore-foot—to optimize energy expenditure across different gait phases. Constant-curvature "arc" feet improve upon flat designs but lack the bio-mimetic complexity required for peak efficiency. This paper introduces the Ellipse-based Segmented Varying Curvature (ESVC) foot, bridging the gap between biological efficiency and mathematical tractability for real-time control.
Design and Methodology
The core innovation of the ESVC foot is its three-part elliptical geometry. Unlike single-curve designs, the ESVC foot features a flatter mid-section for stability and more curved hind and fore-sections to facilitate efficient heel-strike and toe-off.
To integrate this complex shape into a bipedal controller, the authors developed an analytical contact model using elementary functions, avoiding the computationally expensive incomplete elliptic integrals typically required for elliptical arc-length calculations. A Nonlinear Programming (NLP) optimization framework ensures smooth transitions between the three segments by maintaining consistent tangent slopes. Furthermore, a "sine-like" error compensation scheme was implemented to correct approximations in the arc-length model, ensuring high-fidelity center-of-mass (COM) and swing foot trajectory tracking.
Experimental Validation
The design was validated using the TT II (Test Traveller II) biped robot. The robot's control architecture utilized a Hybrid Linear Inverted Pendulum (HLIP) model to manage balance and gait.
The experimental setup compared five distinct foot geometries: a standard Flat Foot, a Line Foot, and three ESVC variants (ESVC1, ESVC3, and ESVC5) with varying eccentricities. The hardware implementation included integrated IMUs, inclinometers, and pressure sensors to monitor contact states and energy consumption in real-time.
Results and Discussion
Physical experiments demonstrated that the ESVC foot significantly enhances walking efficiency. In lateral walking tests at 0.2 m/s, the ESVC foot achieved an 18.52% reduction in average power consumption compared to the line foot. Sagittal walking also saw improvements, with the ESVC5 variant reducing energy costs by approximately 9.68%.
Key insights from the data suggest that "flatter" elliptical profiles (such as ESVC5) provide a superior balance between rollover distance and stability compared to highly curved variants. The segmented design proved particularly effective in the coronal plane, where it helped manage the robot's lateral sway more efficiently than traditional designs.
Conclusion
The ESVC foot offers a robust solution for improving bipedal robot efficiency through bio-inspired geometry and efficient analytical modeling. While the current design relies on a brute-force search for optimal elliptical parameters and assumes rigid ground contact, future work will focus on automated optimization frameworks and adapting the model for soft or uneven terrain.
Summary updated