A quadrupedal robot completed a marathon in South Korea on a single battery charge, traversing the demanding course in 4 hours, 19 minutes and 52 seconds. In the end, the device still had enough energy, according to the researchers, to cover about another 25 kilometers.
RAIBO2 participated in the 22nd edition of the Sangju Marathon, accompanying human runners on a particularly challenging course, with 286 meters of elevation change and sections of slippery terrain. The robot maintained an average speed of 2.64 meters per second and consumed 1,280 Wh during the race.
Developed by a team led by Jemin Hwangbo, the RAIBO2 was designed to reduce the energy consumption associated with locomotion of legged robots. Unlike wheeled robots, these machines must continuously expend energy to support their own weight and compensate for the losses associated with each stride.
To increase autonomy, the researchers combined changes in hardware and software. Among the solutions used are lighter legs, more efficient motor-drive circuits, and movement rules refined through machine learning.
According to the authors, the RAIBO2 achieved a transportation energy cost of 0.25, a metric that relates the energy used during movement to the robot’s weight. This value is lower than 0.37, corresponding to the human value used by the researchers for comparison. The authors consider that it could be the first quadrupedal robot to exhibit a transportation cost below this human reference value.
The RAIBO2’s battery has a total capacity of 2,016 Wh. Based on the energy remaining after the marathon, the team estimates that the robot could have covered another 25 kilometers.
The researchers calculate that the RAIBO2’s range per charge is approximately three times higher than that of other existing quadrupedal robots. The autonomy of these devices has been limited, in particular, by battery capacity, with recent models showing difficulties operating beyond around 20 kilometers.
Increasing autonomy could be relevant for applications in which legged robots must operate for extended periods, notably in search-and-rescue operations in mountainous areas or disaster-affected regions.
The results, published in the journal Nature, provide clues for improving the energy efficiency of legged robots and increasing their operational autonomy.
