Case study
Connected Products and IoT · 2 min read
A production BLE stack for a connected device
A Bluetooth Low Energy stack built for production: an explicit connection state machine, a serialised command queue, fragmentation for large payloads, integrity checks and secure pairing.
- Client
- Connected hardware company
- Industry
- Connected hardware
- Capabilities
- Connected Products and IoT, Mobile Apps
- Status
- Delivered
Highlights
- Connection failure rate reduced in production
- Race-condition GATT failures eliminated by serialising every operation
- Features send and receive whole messages; the protocol stays in one layer
The challenge
Device onboarding kept failing and BLE connections dropped in the field. Every failure became a support ticket, and users who could not connect did not come back. The app also had to exchange payloads of around 10 KB over a link that carries about 20 bytes per write by default.
Our solution
- 01Modelled the connection as an explicit state machine with 12 distinct states, so every event has a defined transition and every failure recovers to a known state.
- 02Sent every GATT operation through one command queue with timeouts, removing the overlapping operations that Android silently drops.
- 03Added a fragmentation and reassembly protocol sized to the negotiated MTU, with checksums so a missing or reordered chunk can never pass as valid data.
- 04Protected pairing against man-in-the-middle attacks, so commands and data cannot be intercepted or injected.
- 05Implemented both the central (Android app) and peripheral roles in Kotlin, so the whole handshake could be tested without waiting for hardware.
What we delivered
- Lifecycle-aware BLE session management with adaptive retry and MTU negotiation
- Priority command queue with a timeout on every operation
- Fragmentation and reassembly over GATT for payloads of around 10 KB
- CRC-16 integrity checks on every message
- Handshake state machine and anti-MITM security layer
- Mapper layer between protocol frames and the Room database
Impact
Connection failure rate reduced in production
Race-condition GATT failures eliminated by serialising every operation
Features send and receive whole messages; the protocol stays in one layer
Technology
KotlinAndroid BLE (GATT)Coroutines and FlowMVIRoomHilt
Have a problem like this?
Start with a free 1-hour architecture review with a senior engineer.