Porting Edge AI Workflows to CHERI/Morello

点击翻译按钮获取中文摘要

Oxon Tech is currently finishing Phase 2 of a GovTech Catalyst funded project on behalf of Mid and West Wales Fire and Rescue Services to create an indoor tracking system, capable of being instantly deployed to incidents with no existing infrastructure requirements.A core component of our platform is a body worn edge IoT device that incorporates an IMU, an ARM process, ARM Microcontrollers and a Neural Processing Unit for running AI The platform also includes a visualisation system that gives incident commanders instant situational overview of where all of their assets are deployed as well as an instantly deployable secure wifi / 4g network.Although the Fire and Rescue Services are an excellent and highly worthy market for this product, commercially the military and law enforcement sectors are far more viable. Creating and securing a platform that spreads across body worn edge iot devices, to incident commander tablets, to cloud based control centers is a huge challenge, and the fact that it is potentially carrying data on troop or police deployment makes it an incredibly likely target for attack.Our main focus will be on porting as much of our existing stack to Morello FVP as possible, identifying and solving any dependency issues that arise. We will also carry out a thorough assessment of options for incorporating hardware devices including NPUs, by either USB passthrough to the FVP if possible or emulation. By the end of the project we aim to have created as much public documentation on porting IoT, ROS based and if time allows FreeBSD based systems as possible, as well as built up significant in-house expertise in DSbD technologies.