Engineers from the Korea Advanced Institute of Advanced Technology (KAIST) organized the participation of the RAIBO2 robot dog in the 22nd Sangju Dried-Persimmon Marathon. The human winner covered the 42 km 195 m distance in 2 hours 36 minutes and 32 seconds, while the RAIBO2 robot completed the task in 4 hours 19 minutes and 52 seconds. The robot successfully completed the course, including two 50-meter climbs, on a single battery charge. However, the complex terrain with elevation differences posed certain challenges, resulting in “unforeseen losses in efficiency.” To solve these problems, reinforcement learning was applied in the RaiSim simulation environment, allowing RAIBO2 to navigate steadily over slopes and icy road sections. A high-torque articulating mechanism helped the robot dog collect excess energy on downhill slopes for use on uphill slopes. Raion Robotics, founded by one of KAIST's professors, developed a vibration and shock mitigation system for RAIBO2 that prevented damage to the mechanics. The company also improved the mechanical and electrical components, as well as the robot's software and artificial intelligence. The authors of the project plan to continue RAIBO2's participation in marathon races, and in the future the robot will have the capability of autonomous navigation. It is expected that such a robot will be able to work effectively in mountainous areas and natural disasters