Skip to main content

device_finder 0.2.0 copy "device_finder: ^0.2.0" to clipboard
device_finder: ^0.2.0 copied to clipboard

Offline Bluetooth device finding for Flutter: a pure-Dart Kalman-filtered RSSI engine with per-observer trilateration, plus a holographic, orbitable 3D radar widget. Pair with device_finder_ble for capture.

0.2.0 #

New: Radar3DView — a holographic, perspective 3D radar.

  • Orbit camera (drag with inertia, double-tap reset, idle auto-orbit), volumetric sweep with a light curtain, sweep pings (orb flash + ground ripple), depth-sorted light pillars and orbs, collision-free one-line labels with a per-size budget, confidence ellipses, range rings for distance-only devices, adaptive "nice" range, floating particle field.
  • RadarTarget (+ fromState / fromStates), Radar3DTheme.dark() / .light(), paintBackground, tap-to-select, semantic summary, reduced motion support, labels use the host app's font.

Engine — correctness and reliability

  • Per-observer filtering: every observerId gets its own fusion window and Kalman filter; anchors are ranged from their own observer. Previously every anchor received the same single-phone distance, so multi-anchor fixes were meaningless. (Behavior change — set observerId = anchor id.)
  • One anchor / phone-only samples now mean distance-only (fix == null) instead of an invented position.
  • Losing anchors keeps the last fix as FixStatus.stale; a stale estimate no longer reports a fresh fix; the proximity trend clears when stale.
  • Mirror resolver: velocity used whole seconds (always 0 at 10 Hz); now uses microsecond timing with smoothed velocity.
  • Name grouping: samples that arrived before setInfo no longer leave an orphan "ghost" device; renaming one member of a shared group keeps the others' data; devices without info expose their own id (never empty).
  • forget() resolves grouped ids and removes stale mappings; new clear(), applyProfile(), reloadProfiles(), deviceCount.
  • tick() notifies at most once per call; the Kalman filter is fed only when new samples arrive.
  • NaN / infinite / absurd RSSI is rejected at the door; duplicate and out-of-order timestamps never move the Kalman state; confidence ellipses can no longer be NaN.
  • Anchor changes are detected by position, not only id.

Tools

  • GeigerHaptics: stop() silences immediately, ticks stop after 3 s without updates, injectable clock/tick for tests, isRunning.
  • FileProfileStore: one corrupt entry no longer discards the rest; writes are serialized. The store is now web-safe (conditional dart:io import) — the package supports web.
  • Stable, platform-identical hashing (exact 32-bit multiply; String.hashCode and plain multiplication differ on the web).

Widgets & example

  • RadarScene keeps an exact fractional phase; MultiRadarView uses a stable bearing hash; DeviceList text ≥ 11 px and no doubled margins; muted text contrast raised to ≥ 4.5:1.
  • New example "Finder" app: simulated room with hubs (no Bluetooth needed) and a Live BLE mode with permission / unavailable states, locate screen with proximity meter, stats, direction and haptics; dark and light themes; adaptive phone / wide layouts.
  • Example fixes: builds from a fresh clone (published device_finder_ble), macOS Bluetooth entitlement + usage string, Android BLE feature optional, timers/subscriptions cancelled on dispose.
  • README screenshots now render on GitHub and pub.dev (previous image URL pointed to a non-existent path).

Requires Flutter 3.29 / Dart 3.7.

0.1.4 #

  • Updated app screenshot in README.

0.1.3 #

  • Fix README screenshot rendering on pub.dev.

0.1.2 #

  • Update README.

0.1.1 #

  • Added app screenshot to README.

0.1.0 #

Initial release.

  • Pure-Dart signal engine: Kalman filter, source fusion, path-loss distance estimation, Gauss–Newton trilateration with mirror-ambiguity resolution, geometric calibration with fit validation.
  • Premium widget catalog: sweeping radar (lit sectors, confidence ellipse), signal meter, device list grouped by broadcast name.
  • Zero-latency repaint architecture (mutate, never rebuild).
  • Geiger haptics for eyes-free search.
  • Offline calibration profiles (JSON file store, atomic writes).
  • Hardware capability adaptation (connected-link preference, TxPower priors).
1
likes
140
points
213
downloads
screenshot

Documentation

Documentation
API reference

Publisher

unverified uploader

Weekly Downloads

Offline Bluetooth device finding for Flutter: a pure-Dart Kalman-filtered RSSI engine with per-observer trilateration, plus a holographic, orbitable 3D radar widget. Pair with device_finder_ble for capture.

Repository (GitHub)
View/report issues

Topics

#bluetooth #rssi #tracking #radar #ble

License

MIT (license)

Dependencies

flutter

More

Packages that depend on device_finder