Outdoor Parking Robots Reach the Curb

The congress slot is an exam, not a contract

HL Robotics said on 16 October 2026 that it will unveil STAN, which the company calls the first outdoor autonomous parking robot, at the 32nd ITS World Congress in Gangneung, Gangwon Province, from 19 to 23 October. Two units are due to run a full park-and-retrieve sequence in the Gangneung Hockey Center lot, with airports named as the kind of site the demonstration is meant to resemble. Chief executive Kim Yungi said Korean parking robots have been commercialised mainly indoors, that outdoor parking is a large part of the market, and that the company wants the congress to show STAN's outdoor driving and its integrated control system.

Read that as a vendor exam with a date, not as an installed base. No price, cycle time, vehicle weight limit, weather rating, or named paying outdoor customer was disclosed. HL Robotics also said it is developing a combined offer with PARKIE, its indoor parking robot, so that drop-off and retrieval can run across both environments. That second sentence is the strategic one. Indoor systems already have a story in Korea. The open question is whether an outdoor machine can share a control model with them without a human dispatcher in the middle.

A facilities director who only needs to know what happens next week can stop at the calendar: two robots, one hockey-center lot, five public days, airport-style entry and exit as the script. Everyone else should use the debut to decide what they would demand before a site trial of their own.

What an outdoor machine has to do that an indoor one does not

Indoor parking robots work in a box. The slab is dry, the light is constant, the bays are painted, and a stalled unit blocks a private aisle rather than a public road. Outdoor work breaks each of those assumptions. Rain changes friction under a drive wheel and under the car. Low sun fools cameras that were tuned under a roof. A painted line disappears under a tyre mark. A pedestrian cut-through appears on a Saturday and not on the survey drawing.

STAN's described control stack is an attempt to manage that mess with a digital twin. In the company's account, the system rebuilds the real bays, shows where robots and vehicles are, shows their routes and job status, sequences work to avoid collisions, and pre-positions cars that are due to leave. Those are the right objects for a control room. They are not, by themselves, a performance claim. A twin can be a live supervisor or a slide. The Hockey Center demonstration will not settle which, unless observers can see a mismatch between the screen and the lot and watch the system respond.

The indoor-outdoor pair matters for a second reason. Many sites that want parking automation are not pure decks or pure fields. A hospital, a mall, and an airport each have a covered handover and an open yard. If PARKIE and STAN cannot pass a car and a job record across that boundary, the operator has bought two products and a clipboard. HL Robotics has said it plans that integrated offer. It has not said the offer is on sale.

A test sheet for the five public days

Procurement and facilities teams attending Gangneung, or watching footage later, can score the debut without waiting for a brochure. Eight checks are enough. None of them requires a private data room.

1. Vehicle class. Note the heaviest and lowest car the robot actually moves. A demonstration sedan is not a loaded van. 2. Entry and exit, both ways. A park without a retrieve is half a product. Time each, from the moment the driver leaves the car to the moment the bay is clear, and again on the way out. 3. Intervention. Count every time a person with a radio steps in, pauses the unit, or repositions a car by hand. One intervention in twenty cycles is a different product from one in three. 4. Conflict. Watch what happens when a pedestrian crosses, when a second robot is moving, and when a car arrives off the script. Sequencing claims are only visible under conflict. 5. Twin fidelity. If a screen is shown, compare it with the lot. A delayed icon is a monitoring tool. A wrong icon is a risk. 6. Surface and weather. Record the pavement, the standing water if any, and the light. Do not generalise a dry afternoon to a coastal winter. 7. Recovery. If a unit stops, time the restart and note whether the car is left blocking an aisle. 8. Handover story. Ask, in public, whether a job started by an indoor unit can be finished by an outdoor unit in this build, or only in a later release.

A team that comes home with those eight notes has something a press statement cannot supply. A team that comes home with a slogan about outdoor autonomy does not.

How this differs from the apartment-robot bill

Korea is already writing rules for robots that share buildings with residents. A draft special act on mobile robots, covered by this publication on 29 September 2026, would let approved buildings run delivery, security, cleaning, and parking robots, and it would ease space standards for robot parking. That legal track and this product track are easy to mash together. They should not be.

The bill, if passed on the timetable described then, is about permission and building design: who may approve a robot, how it calls a lift, and how floor-area rules treat robot parking. STAN's congress debut is about a machine in an open lot. Permission without a working outdoor unit does not move cars. A working outdoor unit without a rule for pedestrian mixing does not scale past a closed demonstration. Buyers need both, on different dates, from different owners.

There is a further split inside the product. A mechanical parking stack, of the kind used for years in dense Asian cities, stores cars in a fixed steel structure. It is slow to install and easy to certify because the motion is on rails. A free-ranging parking robot is the opposite trade: faster to trial on an existing lot, harder to certify once people walk through the aisle. STAN, as described, is the second type. Scoring it against a stack system's theoretical density will flatter or punish it for the wrong reason. Score it against the labour and delay of today's valet or shuttle lot.

Money, without invented prices

No price was disclosed for STAN or for the combined PARKIE offer, so any payback model that starts with a unit cost is fiction. Facilities teams can still build the demand side, which is the side they actually know.

Count staffed hours on the outdoor lot for a normal week and a peak week, including the shuttle or valet overtime that never sits in the robotics budget. Count the bays lost to circulation that a tighter robot layout might recover, and the bays a safety buffer might remove. Count the complaints that come from wait time at departures, because pre-positioning cars is the one feature Kim highlighted that maps directly to a passenger promise. Then refuse to divide those hours by a guessed robot price. Put the vendor's eventual quote beside that sheet. If the quote arrives without a cycle time measured on a lot like yours, it is not yet a quote you can sign.

Airport-style sites deserve a separate line. The company cited airports as the real-world setting the Hockey Center demo is meant to evoke. Airports add kerbside rules, rental fleets, and security boundaries that a hockey-center lot does not have. A clean demo there supports a conversation with an airport operator. It does not support an airport tender response.

The claim that needs a qualifier

"World's first outdoor autonomous parking robot" is the company's wording, reported as the company's wording. It is a marketing superlative until an independent survey of earlier outdoor systems is on the table. Buyers do not need that survey to make a decision. They need to know whether this unit moves their vehicle, on their surface, at a wait time their customers will accept, with a person nearby at a frequency they can staff. First-in-the-world does none of that work.

The other overclaim to retire early is commercialisation by announcement. Kim's comment distinguishes indoor robots, which he said have been introduced and commercialised in Korea, from outdoor parking, which he called a major segment still to be won. That distinction is the honest one. Keep it. A five-day public demonstration is how a company asks the market to look. It is not how a company proves a fleet.

What to put in the trial clause

If a site visit after Gangneung goes well enough to justify a paid trial, the clause can be short. Name the lot, the vehicle classes, the hours, and the weather window. Require the eight checks above as weekly figures, not as a highlight reel. Require a written rule for pedestrians and for a robot that stops under a car. Require the vendor to state whether indoor-outdoor job handover is in this release. Set a kill number: if interventions exceed an agreed rate, the trial ends without a purchase obligation. Leave unit price open until those figures exist.

That is a duller document than a congress unveil. It is also the document that keeps a facilities team from buying a demonstration twice, once in applause and once in a contract.

The Gangneung lot will show whether two STAN units can perform a scripted outdoor park-and-retrieve in public. The buying decision sits one step further out, in the unscripted hour, and it should wait for that hour.

Hero image credit: Editorial illustration generated for this package. It does not depict STAN, PARKIE, Gangneung Hockey Center, or a commercial deployment.

This analysis is based on company statements and public market activity. Figures not stated by the company are excluded rather than estimated.