AI-rewritten: This is a summary of an article from Singularity Hub, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

A robot dog named RAIBO2 recently completed a marathon in South Korea without stopping to recharge, covering over 40 miles on a single charge. This distance is roughly three times farther than other current robot designs. The event took place at the Sangju Dried Persimmon Marathon, where the machine finished in four hours, 19 minutes, and 52 seconds, slightly faster than the average human runner. Despite this impressive speed, the primary achievement was endurance; the robot used only 66 percent of its battery power by the end, leaving it with enough charge to travel approximately 40 miles total.

Researchers from the Korea Advanced Institute of Science and Technology developed RAIBO2 using a holistic approach that mapped energy losses across all major joints and circuits. Previous attempts to improve quadruped robots focused on individual components like mechanical joints or gait control, but this team also utilized an AI-based learning strategy to reduce slipping and tripping risks. The robot’s previous marathon attempt failed at the 23-mile mark due to battery depletion caused by frequent speed adjustments. To fix this, engineers redesigned motor controllers to lower resistance, reduced the weight of the legs by removing material from the hip and calf, and upgraded the AI brain to reward softer landings and cooler motors.

The team identified two main sources of energy waste: mechanical friction and electrical inefficiency. Motors and electronics accounted for about two-thirds of the losses, while physics-related issues like foot collisions and gear friction made up the rest. By installing a larger battery with 33 percent more capacity and enabling regenerative charging on downhill sections, the robot achieved significant gains. While RAIBO2 was not fully autonomous and required human steering at times, its efficiency is around 32 percent higher than that of an average human runner in the race. The commercial spinoff company, Raion Robots, plans to mass-produce the unit with added waterproof and dustproof features for industrial use.