Glee Parking · Hardware specs

Cheese: a thin, stick-on parking sensor

A 6.7 mm sheet, glued to the middle of a parking space, that tells whether a car is parked on top and reports it over a low-power radio network. No drilling, no camera.

We're sharing the design early so people who know hardware, streets or parking can tell us what we're missing.

Status: early design. Nothing on this page has been built or tested with real cars yet. The figures are design targets and estimates, not measurements. Parts for a first test kit were ordered on November 7, 2026.

297 × 210 × 6.7 mmAn A4 sheet, about 1/4 in thick

Two magnetometersNo camera, no radar, no microphone

LoRaWAN radioReports when a space changes, plus an hourly check-in

~5-year batteryEstimate, from two flat 3 V cells

Glued downRoad-marker adhesive, no drilling

~$45–55 eachEstimated parts cost for a small batch

Side view to scale: the LAS7 sensor we use today is 38 mm tall; Cheese is 6.7 mm tall and much wider and flatter. — Side view, to scale, next to the off-the-shelf sensor we use today.

How it works

Cheese sits in the middle of the space, between the wheel tracks, with its long side pointing toward the curb. A parked car is a large mass of steel, and it changes the magnetic field around it.

There's a magnetometer at each end of the sheet, 210 mm apart. A car parked on top changes the field differently at the two ends because it's so close. A car in the next space changes both ends about equally. So Cheese calls a space occupied only when the field changes and the two ends disagree.

It reads the field every 3 seconds.
It decides "free" or "occupied" itself. Each 11-byte message holds that answer, the size of the field change at each end, its battery level and its temperature.
It transmits only when the state changes, plus a short check-in every hour.
Target: an arrival shows on the map in about 10–15 seconds.

Top view of a parallel parking space about 6 by 2.4 metres with a car in it. Cheese is a small rectangle in the middle, between the four tires, with its two sensors marked. — Placement in a parallel parking space, seen from above. The red dots are the two magnetometers.

What's inside

Cheese with its lid off: a black tray full of support posts, a thin circuit board running end to end with a magnetometer at each end, the radio module and a flat antenna in the middle, and two flat battery cells. — Lid off. Parts are shown at their real sizes.

The custom circuit board: a long, narrow strip with a magnetometer at each end and a wider hub in the middle holding the radio module, battery pads and programming socket. — The custom circuit board: one long strip with a magnetometer at each end (A and B) and the radio in the middle.

Physical

Size	297 × 210 × 6.7 mm. About 8.5–9 mm on the road, including the adhesive
Edges	Sloped 1:4, so wheels, plows and street-sweeper brooms ride up and over
Shell	3D-printed black polycarbonate blend. 1.5 mm lid on 7 mm posts every 15 mm, so a tire's load goes into the posts, not the parts. 1.0 mm floor, 4.2 mm cavity
Top surface	Black epoxy with aluminum-oxide grit for grip when wet
Screws	Brass or nylon, so no magnetic metal sits near the sensors

Sensing

Sensors	2 × MEMSIC MMC5603NJ magnetometers, 210 mm apart
Detection rule	Field change of at least 5 µT at either end, and a difference between the ends of at least 2 µT. These starting values come from simulations and will be tuned with real cars
Not included	No camera, microphone, radar or license plate reader. It can't tell which car is parked

Radio

Module	RAKwireless RAK3172-T (LoRaWAN, US 902–928 MHz), rated to −40 °C
Antenna	Molex 211140 flat 915 MHz antenna, 0.1 mm thick
Transmit power	20 dBm (100 mW) or less
Range	About 200–400 m to a gateway with a car parked on top, from a model. To be measured
Network	The Things Network, through our own gateway

Power

Battery	2 × CP355050 lithium manganese dioxide cells: 3.0 V, 2,000 mAh each, 3.5 mm thick, rated −20 to +60 °C. Not rechargeable
Average draw	About 15–25 µA estimated. Target: under 60 µA
Life	About 5 years, estimated. We'll measure the real draw on the first prototype
Regulator	None: the cells power the electronics directly, so nothing wastes current while idle

Electronics

Circuit board	Custom 2-layer board: a 232 × 12 mm strip with a 44 × 30 mm hub, 0.8 mm thick. Made and assembled by JLCPCB
Extras	Reverse-battery protection, a programming socket, and an LED that blinks only at power-up and when it joins the network

Installation

Adhesive	Butyl road-marker pads with pavement primer, the kind used to glue reflectors to roads
Removal	Pry up with a flat shovel. Butyl pads are sold as coming off without damaging asphalt or concrete; we haven't tested this yet
Time	About 10 minutes per sensor, estimated

Compared with what we use today

Our pilot so far has used the KKM LAS7, an off-the-shelf parking sensor. Cheese is our own design.

	KKM LAS7 (today)	Cheese (planned)
Height	38 mm	6.7 mm
Footprint	175 mm round	297 × 210 mm
Sensing	Radar + one magnetometer	Two magnetometers
Mounting	Screws into drilled holes, per the manual	Adhesive, no drilling
Battery	8,500 mAh, 8 years claimed	4,000 mAh, ~5 years estimated
Arrival detected	About 1 minute, per the manual	About 10–15 s, target

Cost

	Per sensor
Small batch (5–6)	About $45–55, plus about $60–70 one-time setup for the circuit boards
1,000 units	About $40, landed and ready to install
10,000 units	About $30, landed and ready to install

These are estimates, not quotes. The biggest one-time costs at volume would be injection molds (about $15–40k) and an FCC test for the finished product (about $3–5k). The radio module itself is already certified. Gateways are separate, about $200 each, and each one covers many spaces.

Build plan

1. Test kit Parts ordered Nov 7

Off-the-shelf boards that click together without soldering, in a 3D-printed tray. It runs the same radio chip and software as the final design, so it answers the two big questions cheaply: can two magnetometers without radar detect cars reliably, and does the radio get out from under a car?

The test kit: an orange 3D-printed tray holding a radio board, a flat antenna and two small magnetometer boards at opposite ends, connected by cables. — The test kit in its printed tray.

2. The real sensor Designed, waiting on the kit

The custom board and the 6.7 mm sheet described on this page. The board design is finished and the order files are ready. We'll only order it if the test kit shows:

At least 95% of arrivals and departures caught over at least 20 real parking events, with no false "occupied" from a car in the next space.
Messages still getting through with a car parked on top.
Real numbers to set the detection thresholds.

Exploded view of Cheese: the black lid lifted above the base, showing the magnetometers at each end and the battery cells in the middle. — Lid lifted off the base.

Where we'd love your feedback

Radio from under a car. We expect a parked car to block 10–20 dB of signal. Is the antenna in the right place, or should it sit at the curb-side edge?
Magnetometers without radar. Will two sensors 210 mm apart reliably tell a car on top from one in the next space, including trucks, buses and cars parked off-center?
Durability. Will a 6.7 mm printed sheet survive tires, plows and street sweepers? What material or structure would you use?
Adhesive in the cold. We found no published strength numbers for butyl pads on asphalt at around 5 °C. Have you seen what holds up?
Heat. The cells are rated to +60 °C, and asphalt in summer can get hotter than that. How would you handle it?
Thinner batteries. The cells set the height. Flatter cells (1–2 mm) hold less energy. Do you know of good options or suppliers?
Installing it for real. What would a city or a parking lot operator need to see before letting us install these?

Anything else you notice is welcome too.

Send us feedback

Cheese is designed by Albert Sveen, with Max Lightfield and William Tang, students at Harvard College.

Email us your feedback