GOYANG — In front of Myeongneung at the Seooreung Royal Tombs in Goyang, Gyeonggi Province, on the 25th, an AI-powered self-driving patrol robot called "Sullabot" rolled without hesitation over a dirt path strewn with rocks. It passed easily over stretches pitted and puddled after rainfall, and kept its pace even as the surface changed from the dirt path to the thick coir matting laid inside the tomb grounds. On this day, the Royal Palaces and Tombs Bureau of the Korea Heritage Service conducted a demonstration of the AI patrol robot at Seooreung as part of an effort to build a smart security system for national heritage sites. At the site, Sullabot drove autonomously along a roughly 200-meter stretch running from Myeongneung to the ritual house, showing off its rough-terrain performance and its ability to detect abnormal situations. The full demonstration route runs about 2.6 kilometers, and a single patrol takes about 50 minutes.
The demonstration is part of the Ministry of SMEs and Startups' "AI Robot Startup Public Demonstration and Procurement Project." Under the program, public institutions make priority purchases of AI robots developed by startups to test their viability in the field. At Seooreung, the robot deployed was the "LKR-T1," an indoor-outdoor patrol robot for irregular environments developed by the robotics company LK Robotics. The robot was built with enhanced rough-terrain capability, in particular to account for Seooreung's terrain, which has many slopes and uneven surfaces. Shaped like a two-legged walker, the robot can keep its balance by moving the opposite leg and joints even when one wheel drops into a hollow or the ground height differs on either side. "The LKR-T1 can climb over steps up to 20 centimeters high," said Kim Seung-hun, CEO of LK Robotics. "It can also run without trouble on the roughly 30-degree slope that runs from Gyeongneung to Daebinmyo within the demonstration route."
The site demonstration also included responses to simulated abnormal situations, such as a fire or a visitor collapsing. When Sullabot detected a fire, a warning message played: "A fire has been detected. To prevent the fire, please immediately stop smoking and using open flames." The thermal-imaging camera mounted on the robot detects fires based on temperature changes, allowing it to catch warning signs even when flames are hidden behind branches and not directly visible. When it finds a collapsed visitor, it issues an alert saying, "A visitor has collapsed," and relays the emergency to the control center. When a fire or a collapse is detected, the administrator's screen at the control center also switches immediately to an alert state. After checking the situation on the ground, an administrator can remotely control the robot to move closer to the location or to resume its patrol.
If the robot encounters a visitor during a patrol, it stops in place to avoid a collision and moves again after the person has passed. Using a 3D lidar sensor, it can detect a collapsed person or an intruder even at night. "We hope to use AI robots to reduce management gaps in conditions where it is hard for people to patrol for long stretches, such as at night or during heat waves," an official at the Korea Heritage Service said. The Korea Heritage Service plans to run the AI patrol robot demonstration through late November, then comprehensively evaluate factors such as driving stability and abnormal-situation detection before deciding whether to make a purchase. Based on the results obtained at Seooreung, it also plans to examine ways to apply AI patrol robots to other national heritage management sites, such as palace and tomb grounds.
