South Korean Navy Tests Humanoid Ship Helmsman
South Korea's Navy tested KAIST's humanoid robot PIBOT as a ship helmsman on July 23, where it received spoken orders, steered a simulated vessel, and reported course changes. According to UPI, researchers used a large language model to support standard ship-handling procedures. The Navy is using the trial to assess maritime human-machine teaming amid reported personnel shortages.
South Korea's Navy tested KAIST's humanoid robot PIBOT as a ship helmsman in a naval bridge simulator on July 23, putting the machine through spoken-command steering tasks under simulated rough weather, nighttime navigation, and close-quarters conditions.
According to AJP, the trial took place at the Naval Education and Training Command's ship-handling training facility in Changwon, in cooperation with the Korea Advanced Institute of Science and Technology. The test marked the Navy's first experiment in which a humanoid robot performed bridge helm duties normally carried out by a crew member, Anadolu Agency reported.
PIBOT received helm orders through a handheld microphone, repeated the command, operated the steering wheel, and reported completion after the simulated vessel reached the directed heading. AJP reported that one order was "Rudder left five degrees, steady on 330." UPI reported that, after stabilizing the vessel on the requested heading, the robot responded, "Steady on course 330."
LLM-supported command execution
UPI reported that the Navy and KAIST researchers used a large language model to enable PIBOT to understand and execute standard ship-handling procedures. The robot's command repeat-back is a familiar bridge-control safeguard: a human helmsman ordinarily confirms the order, adjusts the wheel, and reports when the ordered course is held.
The simulator subjected PIBOT to scenarios including narrow waterways requiring frequent maneuvers, high waves that pushed the simulated vessel off course, and nighttime operations, according to Anadolu Agency and AJP. Anadolu Agency reported that PIBOT executed steering commands across those simulated conditions.
PIBOT is a 165-centimeter, 65-kilogram humanoid with three arms, two legs, and three eyes, according to Chosun. The KAIST team led by professor Shim Hyun-chul originally developed the platform as a pilot robot for aircraft control, Chosun reported. Anadolu Agency reported that its development began in 2022 under an Agency for Defense Development initiative, backed by approximately 5.7 billion won, or $3.8 million, from South Korea's arms procurement agency.
From simulator to sea trials
UPI reported that the Navy has divided the evaluation into four stages:
- •A land-based ship-handling simulator
- •A test aboard a vessel moored in port
- •Daytime navigation at sea
- •Extended operations spanning daytime and nighttime conditions
The Changwon exercise was the first stage. UPI reported that researchers are to assess the time between order receipt and completion, steering accuracy, and usability for naval personnel before moving through the remaining stages.
According to Anadolu Agency, the Navy has framed the work as part of a maritime manned-unmanned teaming system and an effort to reduce sailors' workloads as the pool of military personnel declines. For robotics and ML practitioners, the staged design reflects a broader pattern in embodied-AI deployment: language understanding is only one component, while actuator precision, latency, operator interaction, and safe performance under changing physical conditions require separate validation.
Key Points
- 1The Navy's simulator trial had PIBOT acknowledge, execute, and report helm commands, extending LLM-mediated control into a structured maritime task.
- 2The reported four-stage sequence moves from simulation to moored and sea trials, making latency, steering accuracy, and operator usability measurable gates.
- 3Comparable human-machine teaming programs require robust speech recognition, physical actuation, and fail-safe handoff under degraded operating conditions.
Scoring Rationale
This is an early but concrete application of LLM-supported embodied robotics in a high-consequence maritime setting. The simulator-only result limits immediate operational significance, but the reported progression to vessel and sea tests makes the evaluation relevant to robotics, autonomy, and human-machine teaming practitioners.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
