AutoTach

AutoTach is an Android Auto app that turns your car's infotainment screen into a live, color-coded tachometer. It connects to an OBD-II Bluetooth adapter, reads engine RPM directly off the CAN bus, and renders a full-screen tach on the dashboard display that shifts color in real time as the revs climb. Green at idle, amber through the powerband, red near redline.

Built and tested against the Carista EVO adapter (ELM327 v2.2 compatible), but the parser is generic and works with any standards-compliant ELM327 device.

---

Demonstration

IMG_0388.MOV

IMG_0391.MOV

---

Features

• Android Auto integration via CarAppService + Session + Screen with a NavigationTemplate, so the app renders full-screen on the head unit while the driver is moving — hosted through the Android for Cars App Library (androidx.car.app).
• Custom Canvas rendering to the car's Surface driven by a SurfaceCallback — segmented RPM and speed bars with per-segment HSV interpolation (green → amber → red)
• MVVM — an ObdRepository acts as the apps 'Model' or dat layer, exposing StateFlows that can be read directly on the car service side (since it's not an activity), and read on the phone side through a Viewmodel class that wraps the repository state and methods in an activity-safe manner. Finally the phone Fragments and the Android Auto Screen act as views that simply observe.
• Single-Activity UI with Fragments (MainFragment, ConnectFragment), hosted by a single main activity
• Live OBD-II data parsed from an ELM327 BLE adapter (engine RPM 010C, speed 010D), with multi-frame byte reassembly and full ELM327 error-string handling — see Architecture and The Protocol Stack below.

---

Architecture

The app follows MVVM with a clean three-layer separation underneath the ViewModel. Each layer owns one protocol concern and knows nothing about the layers above it. The phone UI (for setup / pairing) and the Android Auto Screen (for the in-car tach display) are both thin observers of the same ViewModel state:

Head unit ◀── renders ─── CarAppService / Screen ┐
                                                 │  observes
Phone UI ──── observes ───────────────────────── ▶ ViewModel ── owns ──▶ ObdReader ── owns ──▶ BluetoothCommunicator
                                                                       OBD-II / PIDs        BLE GATT + ELM327

• BluetoothCommunicator (transport): owns the BluetoothGatt, runs the GATT callbacks, enables CCCD notifications, sends raw AT command strings, and exposes a suspend fun sendCommand(cmd: String): String. It has no idea what a PID is.
• ObdReader (domain): wraps the communicator and exposes suspend fun getRpm() / getSpeed(). All hex parsing, error detection, and PID formulas live here. It has no idea what GATT is.
• ObdViewModel (presentation): owns the ObdReader, exposes StateFlows for both the phone UI and the Android Auto Screen to observe, and survives configuration changes.

---

Concurrency

Between the Bluetooth connection, the PID-polling loop, and two UIs that redraw on every state change, there's a lot going on off the main thread. Kotlin coroutines and StateFlow are what keep it coordinated.

• StateFlow as the source of truth — RPM, speed, connection status, and the device list are all StateFlows. New subscribers immediately get the latest value.
• repeatOnLifecycle(STARTED) on the phone side, so Fragment collectors pause with the view and never leak past it.
• suspend fun sendCommand wraps the callback-based BLE GATT API in a linear one — the caller writes val response = sendCommand("010C") and awaits the parsed response
• ELM327 init handshake as a coroutine — after notifications are enabled, scope.launch { performHandshake() } sends the 7-command init sequence (ATZ, ATE0, ATL0, …) one after another using sendCommand, so a serial protocol reads as a top-to-bottom loop instead of a callback chain.

---

Tech Stack

• Language: Kotlin
• Android Auto: Android for Cars App Library (androidx.car.app)
• Architecture: MVVM with AndroidViewModel, Fragment, ViewBinding
• Bluetooth: Android BLE GATT API directly — no RxAndroidBle, no Nordic wrapper
• Min SDK: 24 / Target SDK: 35 / JVM: 17

---

Getting Started

1. Clone the repository

git clone https://github.com/hobbbbies/autotach.git
cd autotach

2. Open in Android Studio

Open the project in Android Studio (Giraffe or newer). Gradle sync will pull dependencies.

3. Run on a physical device

BLE doesn't work in the emulator. You'll need a real Android device and an ELM327-compatible OBD-II adapter (Carista EVO, OBDLink, generic clone, etc.) plugged into a car's OBD-II port.

To test the Android Auto experience without leaving the driveway, install the Desktop Head Unit (DHU) from the Android SDK — it simulates a car's head unit on your desktop and renders the in-car UI live.