Researchers at Seoul National University of Science and Technology have developed 3D-printed feet for quadruped robots that reduced battery power consumption by as much as 6.2% in testing. Keun Park and Jung-Yup Kim of the university’s School of Mechanical System Engineering developed porous triply periodic minimal surface, or TPMS, foot structures designed to store energy when a robot’s foot contacts the ground and release it during movement. The researchers paired the feet with a deep reinforcement learning controller intended to coordinate the robot’s gait with the compression and rebound of the structures.
The study, published in the International Journal of Precision Engineering and Manufacturing-Green Technology, tested the approach at walking speeds ranging from 0.4 to 1.0 meters per second. Quadruped robots generally require repeated motor-driven leg movements, making energy consumption a constraint for applications including inspection, transportation and search-and-rescue operations. Elastic components can store energy during foot impact and release it during push-off, but some of that energy can be dissipated at lower walking speeds.
The researchers designed and 3D-printed three hemispherical TPMS foot configurations based on primitive, gyroid and diamond structures. Compression tests were used to assess their energy-storage and release characteristics. A diamond structure with a relative density of 60% was selected for further testing based on its flexibility, energy absorption and energy-loss characteristics.
The researchers then trained a deep reinforcement learning controller to evaluate walking strategies according to their energy consumption while accounting for compression and rebound of the TPMS feet. The controller coordinated the robot’s gait with the passive energy storage and release of the feet to reduce the work required from its motors.
Tests on a commercially available quadruped robot equipped with the optimized TPMS feet showed battery power consumption reductions of between 1.4% and 6.2% compared with conventional solid feet across walking speeds of 0.4 to 1.0 meters per second. The robot maintained stable locomotion during the tests.
“By modulating foot stiffness and leveraging passive energy absorption and release, the proposed approach offers a practical alternative to conventional leg-mounted spring mechanisms,” Park said. TPMS structures consist of repeating three-dimensional porous networks and have been used in applications including airless tires and soft robotic components such as grippers and joints.
“These results demonstrate that TPMS metastructures, when properly modeled and exploited through learning-based control, can serve as effective energy-shaping components for energy-efficient quadruped locomotion,” Kim said. The researchers said the approach could be applied to quadruped robots used for indoor services, inspection and logistics, where reducing power consumption could extend operating time.
