Summary

This document contains details of the release v2022.09 for total-solutions.

Overview

This release adds three main features:

  • ThreadX: Total solutions now support Azure RTOS ThreadX. To compile an application with it use the flag --rtos THREADX on the command line.
  • Azure Certification: The total solution applications can be Azure IoT Cloud certified when DPS provisioning is used.
  • Speech: A new total solution has been added: Speech. It offers the same cloud capability as the Keyword application (AWS and Azure) but runs a different inference model: tiny ASR. This ML model recognise stream of words which are then forwarded to the cloud.

Breaking changes

  • In ats.sh script, keyword should be used instead of kws to build and run the keyword application.

Known issues

  • mlia does not install the CS310 backend as it is broken with the latest AVH. CS300 backend is still installed

Changelog

  • doc: Improve README
  • app: Add speech examples
  • keyword: Remove useless defines.
  • ml: Use latest version of ml kit
  • tf-m: Update partition layout for larger models
  • lib: Add speex to dependencies
  • keyword: Allow override of audio file location
  • keyword: remove audio buffer NS/S alias
  • apps: remove warnings
  • ml: cleanup build system integration
  • ota: move library from keyword into lib
  • bsp: rework config and move aws
  • examples: move examples into their folder
  • ci: Set environment variables for AMI pipeline
  • Updated diagram with speech application and Threadx components
  • Add provisioning to the azure demo
  • rtos: Add ThreadX as an RTOS option
  • Updated architecture diagram
  • Update ML eval kit to version 22.05
  • mlia: Roll back mlia to 0.3.0, to workaround backend install failure
  • sdk: Update SDK and toolchain repos to the latest
  • ci: Create pipeline for AVH AMI testing
  • ci: Move AVH Docker build outside keyword
  • quality: Remove project specific issue and MR templates
  • tf-m: Switch to IPC model which is the default
  • mlia: Update README for MLIA
  • mlia: Save output into file
  • clang-format: Sort includes
  • pre-commit: Ignore lib folder
  • CI: Avoid duplicate pipelines.
  • Azure: Fix test
  • CI: Add AWS cleanup script.
  • CI: Factorization of jobs.
  • AWS: Rework cleanup to resources.
  • CI: Use UUID to generate tests ID.
  • quality: Add default issue and MR templates
  • FreeRTOS: Enable explicitly FPU support
  • KWS: Raise ML task priority
  • Ethos: Re-enable RTOS integration

Testing

Automated CI tests

  • Static checks
    • File formatting checks
    • Clang format checks
    • gitlint
  • Build & Test
    • blinky
    • speech
    • keyword
  • Configurations
    • Targets:
      • Corstone-300
      • Corstone-310
    • RTOS:
      • FreeRTOS
      • CMSIS RTOS2
      • ThreadX
    • Cloud clients:
      • Aws IoT cloud
      • Azure Iot cloud
    • AMI version: 1.3.0