Glee hardware research: what should the sensor be?
Live at research.gleepark.me. This folder is updated continuously by a research loop: hypotheses, experiments (code + data + results) and the current recommendation. Nothing here has been tested on a street yet unless it says so.
The question
What hardware tells a city, in near real time, whether each on-street parking space is free? It must be cheap enough to cover whole streets, easy enough that a city will let us install it, and right often enough that drivers trust the map. The form factor is open. It doesn't have to be "cheese" (the thin sheet stuck under the car).
Breaking it down
| Sub-problem | What decides it | Status |
|---|---|---|
| 1. Sensing: how do we know a car is there? | accuracy, cost of the sensor, power per reading | Experiment 1 done (simulation); method survey in progress |
| 2. Placement / form factor: in the road, on the curb, on a pole, overhead? | permits, install time, winter, theft, what the sensor can "see" from there | deployment survey in progress |
| 3. Power: battery, solar, mains? | where it's mounted (light), what it draws | model done for cheese (hardware/swiss/SCENARIOS.md) |
| 4. Radio / network: LoRaWAN under a car, gateway density | link budget, gateway cost per space | model done; real under-car numbers come from American cheese |
| 5. Cost and scaling: $ per space per year, all-in | unit cost, install labour, maintenance, gateways, network | next experiment |
| 6. The city: what Cambridge will approve and maintain | permits, liability, snow operations, privacy | deployment survey in progress |
Hypotheses
| # | Hypothesis | How we test it | Verdict so far |
|---|---|---|---|
| H1 | Two magnetometers in the road (cheese) can tell a parked car from neighbours and traffic without a radar | Exp 1 (simulation), then American cheese on real cars | Supported in simulation (99.7 %), but independent field studies put magnetometer-only sensors at 81-93 % (sensing report). Real test decides |
| H2 | A magnetometer on the curb (no road work) is good enough | Exp 1 | Rejected: ~90 % in simulation |
| H3 | A small radar makes detection robust | survey + Exp 4 (XM125 radar on real cars) | Strongly supported by the industry: Bosch, Libelium, Option, Fleximodo and KKM all moved to magnetometer + radar. A 60 GHz A121 radar is ~$10, 5 × 5 mm, and adds < 1 µA if it only confirms magnetometer events |
| H4 | Mounting on existing meter poles beats anything in the road for permits, winter and maintenance | deployment report, Exp 3 | Supported for deployment: Bow Street has single-space meters (poles at the curb), no digging, no lane closure, above plows. Open on accuracy: Boston's meter-mounted sensors "had many problems with the data's accuracy" |
| H5 | One sensor can cover several spaces and win on cost per space | Exp 2 | Cameras do, but Cambridge's Surveillance Ordinance makes them a poor fit; a non-imaging pole radar covering 2 spaces is the candidate |
| H6 | The radio signal itself (RSSI) is a free second opinion | American cheese logs | Supported in literature: LoRa RSSI alone reached 96-97 % in one study. Free: log it from day one |
| H7 | Not every space needs a sensor | literature, later data | DDOT found ~50 % coverage + meter-payment data predicted block occupancy. Matters if the product is "free spaces on this block" rather than per-space enforcement |
| H8 | Cheese (stuck to the road) survives a Cambridge winter | deployment report | Rejected for winter: proud of a plowed, swept, salted surface; Bow Street's odd side is a snow tow route. Seasonal pilot device only |
Experiments
| # | Experiment | Result |
|---|---|---|
| 0 | Pilot data: what the two LAS7 pucks reported Nov 2-7 (data) | Bench testing, not street parking (median episode 2.4 min, signal up to -25 dBm). No street data exists yet |
| 1 | Which architecture detects best?: Monte Carlo, 2,500 situations | In-road arrays 99.7-99.9 %, single in-road magnetometer 98.5 %, radars ~99.5 % (assumed), curb magnetometer ~90 % |
| 2 | Cost per space per year, by form factor | next |
| 3 | Pole-mounted radar geometry: own space vs the neighbour's, snowbanks | next |
| 4 | Real radar on real cars: SparkFun XM125 (A121) on American cheese's mux, under a car and on a pole | needs a $50 part: SparkFun SEN-24540 |
Reports
- Every way to detect a parked car: methods, accuracy, power, prices, what vendors use
- Deploying on Bow Street: permits, winter, poles, pilots elsewhere, questions for the city
Current direction
The research points away from "one thin sheet stuck under every car" as the city-scale product. The current direction has three parts:
- Sensing: magnetometer + radar, not magnetometer alone. The magnetometer watches continuously at µA; a 60 GHz pulsed radar (A121) confirms. This is what the industry converged on after magnetometer-only sensors scored 81-93 % in independent tests.
- Form factor: two tracks.
- Cheese (in-road, under the car) stays the learning device: American cheese now, Swiss cheese for the even (non-tow-route) side of Bow Street until late November. It gives us real under-car data. Add the radar to it.
- A meter-pole sensor (working name "Provolone": it hangs on the pole) is the leading city-scale candidate: clamp-on to the existing single-space meter poles, radar looking at the car beside it, no digging, no lane closures, above the plows, big battery (size isn't constrained on a pole). Its open risk is accuracy, which Exp 3 and Exp 4 attack.
- Business: sell occupancy data, not hardware, and fuse it with free signals: radio RSSI, and meter-payment data if the city shares it.
Next
- Exp 2: cost per space per year (unit, install, winter handling, maintenance, gateways) for cheese, cored puck, meter-pole radar.
- Exp 3: pole-radar geometry simulation (beam angle, range gate, neighbour cars, snowbanks).
- Re-fit the car magnetic model when the measured-magnetism report lands.
- Order one XM125 radar ($50) and plan Exp 4 with American cheese: radar under the car and on a pole, alongside the magnetometers, with RSSI logged.
- Questions for the city (from the deployment report): who approves a device on a meter pole; does a non-imaging radar count as surveillance; Bow Street paving moratorium.
Experiment 1: which architecture detects parked cars best?
Monte Carlo, 2500 situations (half occupied). Thresholds are chosen on one half of the trials and scored on the other. Generated by research/sim/architectures.py.
| Architecture | Accuracy | Missed cars | False alarms | Best threshold |
|---|---|---|---|---|
| In-road 4-magnetometer array (24 x 20 cm) | 99.9 % | 0.2 % | 0.0 % | 2.8 uT |
| In-road 2-magnetometer gradiometer (cheese, 21 cm) | 99.7 % | 0.5 % | 0.2 % | 1.1 uT |
| In-road upward radar (A121-class) | 99.6 % | 0.5 % | 0.3 % | — |
| Pole/meter-post side radar or ToF | 99.5 % | 0.4 % | 0.6 % | — |
| In-road single magnetometer | 98.5 % | 1.0 % | 2.1 % | 2.1 uT |
| Curb-face magnetometer | 90.2 % | 9.9 % | 9.8 % | 1.6 uT |
What each trial randomises: car size (small car to van) and where it stops in its 6 m space; gap to the curb (10-50 cm); cars in the spaces in front and behind (70 % each, anywhere in their space); a car or bus in the travel lane (30 %); street direction vs magnetic north; sensor noise (0.2 uT RMS); temperature drift since calibration (0.5 uT shared, plus 0.3 uT per sensor); each sensor's ~1 % gain error on Earth's field; and how magnetic cars really are (log-normal multiplier, x0.4 to x2.5, because our dipole model is uncalibrated).
Assumptions to check against American cheese's data: the dipole car model; radar/ToF modelled as geometry with 99.5 % per-reading reliability (no snow on in-road radar, no snowbank in front of a pole sensor); a single reading per decision (real firmware confirms over 3 readings, which cuts false alarms further for all magnetometer designs).
Report: deploying on Bow Street, Cambridge
Research agent, 2026-11-07. Primary sources linked; [unverified] marks estimates.
Site facts that drive the choice
- Bow Street has single-space meters (about 40 spaces, BOW-0001 to BOW-0057, $2/hr, 8 AM-8 PM, 30 min or 2 h,
SmartMeter = No) in the City's metered-spaces GIS layer (layer, dictionary). Coin + app meters, no replacement planned (Cambridge Day). So there are meter poles at the curb; one pole per space vs twin heads per two spaces is [unverified]. - The odd-numbered side of Bow Street is a primary Snow Emergency tow route (~235 ft); cars are towed during bans and plows run to the curb (Snow Center). Brine pre-treatment since 2018; monthly mechanical sweeping April-December (Street Cleaning).
Form factors
| Form factor | Permit path | Install | Winter | Theft / liability | Maintenance | Power |
|---|---|---|---|---|---|---|
| A. Under-car adhesive sheet (cheese) | no published category; written TPT/DPW consent + space coned while it cures | < 10 min, 1-2 people | highest: proud on a plowed, swept, salted surface; vendors steer snow regions to in-ground (Urbiotica) | low theft; raised edge is a minor trip/bike hazard | must come out before each snow emergency | battery only |
| B. Core-drilled flush puck | DPW Street Opening Permit + bond; no work on streets resurfaced in the last 5 years; winter construction moratorium (DPW regs) | 15-30 min, 2 people, cones, maybe police detail [unverified] | low (Nwave: "designed for heavy snowplough regions") | low; flush | battery death = re-coring; SFpark batteries ~3 yr, some sensors paved over (SFpark eval) | battery only |
| C. Curb face | no precedent; DPW sign-off likely [unverified] | 10-15 min | medium-high: banks bury it, plow/sweeper contact | moderate theft | easy unless buried | battery |
| D. Sidewalk / curb top | DPW sidewalk obstruction / annual use permit | ~10 min | medium | ADA clear width + cane detectability (PROWAG); least likely to be approved | easy | some solar |
| E. Clamp-on meter / sign pole | TPT agreement for meter poles; City light poles need a Pole & Conduit Commission grant of location, "even ... temporary installations" (minutes) | 5-10 min, 1 person, no closure [unverified] | low: above snow, away from blades; banks can block a low-aimed view | moderate theft (tamper-proof clamp; ≤ 4 in protrusion at 27-80 in height) | easy battery swaps. Boston found meter-mounted sensors "had many problems with the data's accuracy" and needed "significant staff time to calibrate" (Boston report) | small solar on top possible but canyon-shaded |
| F. Overhead camera / radar | grant of location + Surveillance Ordinance (Ch. 2.128) for anything camera-like: Council approval after a public hearing; Flock ALPRs were switched off at the Council's request | bucket truck, 1-2 h per pole [unverified]; many spaces per unit | low | low theft; high political risk if a camera | few units, but one failure = many spaces | streetlight AC possible |
Evidence from other pilots
- SFpark (~8,000 sensors): interference varied space to space and nearly half the spaces needed two sensors; batteries failed early and at random; occupancy reliable, turnover/stay length not; SFMTA bought data rather than hardware and later replaced sensors with a model built on payment data.
- Boston (330 Streetline in-ground sensors, 2013): sensors "do not give plug-and-play reliably high-quality occupancy data"; scaling needs methods "other than a sensor at every meter" (New Urban Mechanics).
- No published Boston-area winter damage rates found.
Power
No solar under a car. At curb/pole height in a street canyon, winter sun is weak and often blocked; solar-only curb devices have struggled in northern winters. Primary lithium battery for 5+ years; solar trickle on a pole as a bonus.
Takeaways
- Cheese is a seasonal pilot device, not a city-scale product. If kept: even (non-tow-route) side of Bow Street only, recover before the first snow emergency.
- Clamp-on to the existing meter poles is the most deployable option: no dig, ~5 min, no lane closure, above plows, easy battery swaps, probably one TPT agreement. Its risk is accuracy (Boston's experience): it must be proven, including when snowbanks block the view.
- Core-drilled pucks are the proven winter answer but slow to permit and costly to service: a vendor play, not a student one.
- Avoid cameras (Surveillance Ordinance; Flock episode). Ask whether non-imaging radar counts.
- Design for a primary battery (5+ years) and over-the-air recalibration; report health data.
- Sell the data, not the hardware: SF and Boston both found per-space hardware maintenance was the long-term pain.
Questions for Jeffrey Parenti / TPT
- Who approves a device on a meter pole (TPT or the Pole & Conduit Commission)? Insurance/indemnity?
- Is there a permit category for a surface-adhered device in a space? Any DPW objection on a tow route?
- Is Bow Street a 5-year moratorium street?
- Does a non-imaging radar or magnetometer count as surveillance technology under Ch. 2.128?
- Any planned meter replacement (integrate rather than clamp)?
- Who removes/reinstalls devices before snow emergencies and sweeping?
- Can we get ground-truth (meter / PayByPhone / Passport sessions) to validate accuracy?
Report: every way to detect a parked car
Research agent, 2026-11-07, from datasheets, papers and city reports. † vendor claim, not independently tested. ‡ estimate or not confirmed in the primary document. "SFpark 2014" figures via Table V of arXiv 1902.07401.
| Method | Accuracy in parking use | Avg power at ~3 s | Unit cost @1k | Mounting | Weather risk | Notes |
|---|---|---|---|---|---|---|
| Magnetometer, single | 81 % (StreetSmart, SFpark 2014); 86 % SFpark (Roman et al.); F1 0.93 over 13 months outdoors (Huang, PerCom'17); 99 % in a campus lot (Dartmouth); 98 % Urbiotica† | ~2-10 µA‡ | sensor ~$1; finished pucks $245-299 | in-road / surface | very low; works under snow | a parked car changed the field ~27-37 µT in one AMR study (KSC); failure modes: cars next door, rail/power-line interference, high-clearance vans, aluminium EVs, drift |
| Magnetometer gradiometer (Glee) | no published parking figure | same | ~2× sensor | same | very low | Huang used two magnetometers 95.5 mm apart to tell arrivals from departures |
| 60 GHz pulsed coherent radar (Acconeer A121) | radar-only Sensys MicroRadar 98 % (SFpark 2014) | 0.2 mW at 0.1 Hz, 0.23 mW at 1 Hz → ~30-60 µA at 3 s‡; < 1 µA if it only wakes on a magnetometer change‡ | A121 ~$9-13 (DigiKey) | in-road (up), curb, pole | water film / snow on the window: 2-4 dB per 0.5 mm water, 5-20 dB snow/ice (77 GHz figures) | measures distance to stationary objects (30-600 mm in Option SPS200); chip ~5×5 mm: fits a thin sheet |
| 60 GHz Doppler (Infineon BGT60LTR11) | not suitable alone | 2.3 mW min | ~$5-10‡ | any | as above | motion only: can't see a car that's already parked |
| 24 / 77 GHz radar | mag+radar Fybr 78 % (SFpark 2014) | mA-level‡ | $10-30‡ | pole / overhead | as above | overkill for one space |
| Optical ToF (VL53L1X) | IR (CPT) 92 % (SFpark 2014) | ~25 µA at 3 s‡ | $2.5-2.9 | in-road / curb | high: dirt, snow, leaves, sunlight | Urbiotica: a dirty IR sensor reports "occupied" |
| Passive light (shadow) | low alone; light + mag + RSSI fusion F1 0.98 (Huang) | ~1 µA | < $1 | in-road | clouds/trees | good wake-up trigger |
| Ultrasonic | up to 99.9 %†, mostly indoor | mA per ping‡ | $20-100 | overhead / curb | wind, rain, snow, temperature | garages |
| Camera (pole / overhead) | Parquery 99 %†; Cysen 77 % (SFpark); Bellevue best vendor ~83 % | watts | ~$32 per space (MIT est.) | light pole / building | night, glare, snow, occlusion by trucks | Cambridge Surveillance Ordinance; Flock cameras shut off in 2025 |
| Inductive loop / piezo | high for moving traffic | mains | ~$350 per loop (1980s $) | saw-cut | pavement damage | not practical per space |
| RF RSSI (LoRa/BLE) | LoRa RSSI 96 % (HMM) / 97 % (NN) (Dujić Rodić); Libelium mag + RSSI 99 %† | ~0 extra | $0 | any | rain, people, multipath | separates own car from neighbours only when baseline RSSI > -90 dBm (Huang) |
| Wi-Fi sensing / acoustic | research stage, not per space | not pursued |
Commercial products
| Product | Uses | Accuracy / price |
|---|---|---|
| Bosch Parking Lot Sensor | magnetometer + radar | 96 %†; 5 yr† (integrator expects 2-3) |
| Libelium | mag → mag + cloud RSSI (2018) → mag + radar (2019) | 99 %†; radar added against interference, neighbours, double-parked cars |
| Option Spot-On | SPS100 → SPS200 adds 60 GHz pulsed coherent radar | 10 yr on 8100 mAh at ≤ 50 events/day† |
| Fleximodo | mag + radar | ~95 % mag-only vs 99.96 % mag + radar† |
| Nedap SENSIT | IR + mag | "close to 100 %"† |
| Urbiotica | dropped mag + IR for mag-only | 98 %†; IR fails when dirty |
| PNI PlacePod | RM3100 magnetometer | $290-299 retail |
| Sensys | magnetometer, MicroRadar variant | MicroRadar 98 % (SFpark) |
| Fybr | mag + radar | 78 % (SFpark) |
| Cleverciti | overhead optical | up to 100 spaces per unit; Redwood City $904k + ~$18k/month |
| IPS (meter / pole mount) | stereoscopic + magnetometer | Santa Ana: $250/sensor + $45 install + $3.50/space/month |
| KKM LAS7 | geomagnetic + microwave radar | no datasheet found |
City prices: NYC $250 per sensor (2 per space); Milwaukee $245; Burlington VT $250.
Takeaways
- The industry moved from magnetometer-only to magnetometer + radar (Bosch, Libelium, Option, Fleximodo, KKM): the magnetometer watches at µA, the radar fires only on a change to confirm. Urbiotica is the exception (mag-only, 98 %†).
- Independent long-term data puts magnetometer-only at ~81-93 %, not 98-99 %. Glee's target should be a ground-truthed, multi-month measurement on Bow Street.
- Best upgrade for cheese: a 60 GHz pulsed coherent radar (A121, ~$10, ~5×5 mm) that only confirms magnetometer events (< 1 µA extra). Avoid Doppler-only parts. Risk: slush/water on the window (hydrophobic window; ignore returns closer than ~5 cm).
- RF attenuation is free: every TTN uplink carries RSSI/SNR; LoRa RSSI alone reached 96-97 %. Log it now (works best when baseline RSSI > -90 dBm).
- ToF/IR get dirty in the road; better on a curb or pole. Cameras: cheapest per space but a poor fit for Cambridge (ordinance, power, 77-83 % in curb pilots).
- Not every space may need a sensor: DDOT found ~50 % sensor coverage fused with meter-payment data still predicted block-level occupancy (DDOT).