Shipping Flutter + Smart Glasses + Jetson + ROS 2: An End-to-End Checklist

dev.to2026-11-05 23:140 views

The problem

Cross-device systems fail at boundaries: mismatched SDK versions,
unsupported hardware, permissions, unstable Wi-Fi, overloaded inference,
unhandled robot errors, or missing observability. A release checklist
makes these risks explicit.

What this guide covers

This tutorial focuses on practical architecture and performance
engineering for production systems. The central principle is to measure
the complete pipeline, keep latency-sensitive work close to the
hardware, and make failure states visible to the operator.

Reference architecture

Flutter client ↔ secure API/gateway ↔ glasses adapter and/or ROS 2
services ↔ Jetson inference/model runtime ↔ robot controller. Each
boundary has a versioned contract, timeout, health signal, and failure
policy.

Implementation plan

Create a compatibility matrix for Flutter, Android, wearable SDK, ROS 2, JetPack, CUDA/TensorRT, and model versions.
Define latency budgets for UI response, camera frame age, inference, command acknowledgement, and telemetry freshness.
Verify permissions, encryption, authentication, least privilege, retention, and data redaction.
Run fault injection for device disconnect, app process death, service restart, packet loss, overload, and invalid model output.
Document rollback, known limitations, hardware setup, logs to collect, and safe recovery steps.

Example

release_gates:\n - startup_and_reconnect_tested: true\n - p95_latency_within_budget: true\n - stale_data_is_visible: true\n - robot_commands_are_authorized: true\n - thermal_soak_test_passed: true\n - rollback_artifact_available: true

The snippets are illustrative starting points, not drop-in integrations.
Match APIs and version requirements to the exact SDK, OS image, hardware
revision, and runtime in your project.

Performance checklist

Measure end-to-end latency, not only one function or model invocation.
Report p50 and p95/p99 latency alongside throughput, memory use, and dropped work.
Bound queues and define what happens when consumers fall behind.
Avoid repeated frame copies and expensive work on Flutter's UI isolate.
Test sustained workloads after the device reaches thermal equilibrium.
Keep UI state, device state, and robot state distinct.
Add timeouts, cancellation, reconnection, and explicit stale-data indicators.
Profile on the actual target hardware using a reproducible build.

Common pitfalls

Do not treat a green unit-test suite as proof of physical-system safety.
Do not ship undocumented API assumptions or unpinned model/runtime artifacts.
Do not expose diagnostic endpoints or debug credentials in production builds.

Hands-on exercise

Run a simulated release review with a failure injected at each boundary.
Record detection time, user-facing behavior, recovery action, and
regression test.