Comparison between External and Internal Single Stage Planetary gearbox actuators for legged robots

TL;DR

This paper systematically compares internal and external single-stage planetary gearboxes for legged robot actuators, providing a design framework and demonstrating the optimal choice depends on gear ratio, with experimental validation.

Contribution

It introduces a systematic design framework for selecting planetary gearbox architectures based on performance needs and validates it through optimized actuator prototypes.

Findings

ISSPG is lighter at gear ratios 5:1 to 7:1.

ESSPG becomes preferable above 7:1 gear ratio.

Optimized prototypes match predicted masses closely.

Abstract

Legged robots, such as quadrupeds and humanoids, require high-performance actuators for efficient locomotion. Quasi-Direct-Drive (QDD) actuators with single-stage planetary gearboxes offer low inertia, high efficiency, and transparency. Among planetary gearbox architectures, Internal (ISSPG) and External Single-Stage Planetary Gearbox (ESSPG) are the two predominant designs. While ISSPG is often preferred for its compactness and high torque density at certain gear ratios, no objective comparison between the two architectures exists. Additionally, existing designs rely on heuristics rather than systematic optimization. This paper presents a design framework for optimally selecting actuator parameters based on given performance requirements and motor specifications. Using this framework, we generate and analyze various optimized gearbox designs for both architectures. Our results demonstrate that for the T-motor U12, ISSPG is the superior choice within the lower gear ratio range of 5:1 to 7:1, offering a lighter design. However, for gear ratios exceeding 7:1, ISSPG becomes infeasible, making ESSPG the better option in the 7:1 to 11:1 range. To validate our approach, we designed and optimized two actuators for manufacturing: an ISSPG with a 6.0:1 gear ratio and an ESSPG with a 7.2:1 gear ratio. Their respective masses closely align with our optimization model predictions, confirming the effectiveness of our methodology.