PROJECT / 05 FEATURE ARTICLE
SignalAtlas
An offline, read-only field instrument for observing and interpreting nearby Wi-Fi and Bluetooth LE broadcasts.
01 Project overview
A field guide for nearby radio broadcasts
Signal Atlas runs on the FoloToy AI Passport and observes nearby Wi-Fi access points and Bluetooth LE advertisements. It turns bounded radio metadata into local, human-readable information across Nearby, Detail, and History, with ALL, WI-FI, and BLE viewing modes.
The instrument is offline and read-only. It describes broadcasts that are already present; it does not connect to networks or devices, interrogate them, or upload observations.
02 Observation pipeline
One observation model for Wi-Fi, BLE, and simulation
Simulator Wi-Fi and BLE observations are mock input through the shared application pipeline; they do not establish real radio behavior.
03 Engineering choice
Observe broadcasts without interrogating devices
CHOSEN
Passive, read-only observation
NOT CHOSEN
Connecting to or interrogating discovered devices
Signal Atlas interprets information that nearby devices already broadcast instead of initiating Wi-Fi association, BLE connections, GATT interrogation, or active probing.
No Wi-Fi association, BLE/GATT connection, client sniffing, monitor-mode product path, RSSI distance estimates, attack detection, or security scores; rotating BLE private addresses are not correlated into guessed physical identities.
04 Field workflow
Scan, inspect, remember
- 01
Nearby
Observe the current bounded list of Wi-Fi and BLE broadcasts.
- 02
Choose mode
Switch among ALL, WI-FI, and BLE to choose which observations to view; use the Quick Menu for field controls.
- 03
Inspect Detail
Open an observation and distinguish OBSERVED facts from deterministic RESOLVED information and any explicitly bounded POSSIBLE inference.
- 04
Review History
Revisit locally stored identity and sighting summaries while keeping rotating BLE private addresses distinct.
- 05
Pause or resume
Pause scanning while preserving the current view and state, then resume observation; changing signal updates should not make the current selection jump unpredictably.
05 Implementation
Bounded for an ESP32-C3
The target is the FoloToy AI Passport: an ESP32-C3 with 8 MB Flash, no PSRAM, a 240 × 320 display, and physical UP, DOWN, and OK controls. Bounded-memory architecture uses fixed-capacity stores and queues, with bounded alternating Wi-Fi/BLE scan scheduling.
Wi-Fi, passive NimBLE, and Simulator inputs use a shared normalized RadioObservation model. Wi-Fi identity uses BSSID; BLE identity uses address and address type, without correlating rotating private addresses into guessed physical devices. Offline registry tables resolve Bluetooth Company IDs, Service UUIDs, Appearance values, iBeacon, and Eddystone UID, URL, and TLM data. NVS keeps local history. The device UI uses the Instrument Green visual system and Mini Radar scan indicator.
Nordic Assigned Numbers data is included under its documented BSD-3-Clause terms. Public source and default firmware do not redistribute IEEE MA-L, MA-M, or MA-S derived assignment records. Tooling retains a path to generate local tables from independently obtained official IEEE data; this does not imply IEEE redistribution permission.
06 Verification
Reviewed pre-hardware state
REVIEWED STATE
- Host / static
- PASS
- Full validation
- PASS
- ASan / UBSan
- PASS
- Stress
- 50 cycles PASS
- Production Full Flash
- 1,716,912 B · PASS
- Simulator Full Flash
- 1,571,024 B · PASS
- Simulator
- PASS
- Physical RF
- UNVERIFIED — REAL DEVICE REQUIRED
Simulator validation covers supported software, UI, and shared-pipeline paths. Real BLE RF and authoritative Wi-Fi RF, the physical display and controls, real-device memory, Wi-Fi/BLE coexistence, scan density and range, repeated radio lifecycle, and long-running RF stability remain UNVERIFIED — REAL DEVICE REQUIRED.