Insider Brief
- SeoulTech researchers developed 3D-printed footpads and an AI controller that reduced a quadruped robot’s battery use by as much as 6.2% while maintaining stable movement.
- The porous diamond-pattern footpads absorb impact energy when the robot steps and release it during push-off, while deep reinforcement learning adjusts the gait to reduce motor effort.
- Tests showed power savings of 1.4% to 6.2% at walking speeds of 0.4 to 1 meter per second, suggesting potential applications in inspection, logistics, indoor services and search-and-rescue operations.
Researchers at Seoul National University of Science and Technology have developed 3D-printed footpads that helped a quadruped robot use as much as 6.2% less battery power, pointing to a relatively simple way to extend operating time without adding springs or other hardware to its legs.
According to the university, also known as SeoulTech, the approach combines flexible, porous footpads with an artificial-intelligence controller that learns how to use the energy absorbed when a robot’s foot strikes the ground. The controller times the robot’s gait so that some of that stored energy is released during the next step, reducing the work required from its motors.
The study was published online May 5 in the International Journal of Precision Engineering and Manufacturing-Green Technology. The research was led by Jung-Yup Kim and Keun Park of SeoulTech.
According to the researchers, robots equipped with the optimized footpads consumed between 1.4% and 6.2% less battery power than robots using conventional solid feet while walking at speeds ranging from 0.4 to 1 meter per second. Importantly, the machines remained stable during the tests.
The researchers noted that the results could be useful for quadruped robots used in inspection, logistics, indoor services and search-and-rescue operations, where longer battery life can increase the distance a machine travels or reduce the frequency of charging and battery replacement.
Four-legged robots generally require more energy than wheeled machines because they must repeatedly lift, move and place their legs. Engineers have tried to recover some of that energy by adding springs or other elastic parts to robot limbs, but those systems can lose much of the stored energy at lower speeds and may make the robot harder to control.
The SeoulTech team instead placed the energy-absorbing structure in the feet.
Testing Porous Foot Designs
The researchers used structures known as triply periodic minimal surfaces, or TPMS, that are lightweight materials made from repeating three-dimensional patterns containing networks of open space. Such structures can flex under pressure, absorb an impact and then return toward their original shape. They have also been explored for airless tires, soft robotic grippers and flexible joints, researchers pointed out.
The team designed and 3D-printed three rounded footpads using different internal patterns known as primitive, gyroid and diamond structures, and then compressed the pads to measure how much energy they absorbed, how much they returned and how much was lost.
The researchers selected a diamond-pattern design with a relative density of 60%. According to the study, that version provided the best balance of flexibility, impact absorption and limited energy loss among the designs tested.
A softer foot alone, however, was not enough to produce the full efficiency gain. The robot also needed to adjust its movements to match the way the footpad compressed and rebounded.
“By modulating foot stiffness and leveraging passive energy absorption and release, the proposed approach offers a practical alternative to conventional leg-mounted spring mechanisms,” noted Park.
Teaching the Robot to Use Stored Energy
The researchers trained a controller using deep reinforcement learning, a method in which an AI system improves by testing different actions and receiving feedback on the results. In this case, the controller evaluated walking strategies based partly on their energy use. It learned to coordinate the robot’s gait with the timing of the footpads’ compression and recovery.
That reduced the amount of force the motors needed to generate and limited unnecessary corrective movements, according to SeoulTech. The controller effectively treated the flexible footpads as passive energy-storage devices rather than as simple cushions.
“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,” added Kim.
