Cyber-Physical AI on Embedded/Edge/Cloud Architectures.

  • Cyber-Physical AI on Embedded/Edge/Cloud Architectures.

School of Electronics, Electrical Engineering and Computer Science
& ECIT Global Research Institute

Proposed Project Title: Cyber-Physical AI on Embedded/Edge/Cloud Architectures.

Principal Supervisor: Dr. J. McAllister

Project Description:

Cyber-physical computing architectures are undergoing undergoing a revolution, evolving into hybrid cyber-physical embedded devices, such as robots, drones, cameras or vehicles which interact with the physical world, along with edge analytics servers and high-performance computing resources in the cloud for advanced data analytics and AI. The problem of designing these systems involves realising a given workload (such as machine vision, recognition and resulting actions) on highly-distributed platforms, to meet a set of performance, speed and energy expenditure requirements.

Critical to being able to solve this problem is being able to productively program these hybrid platforms. This is currently a critical problem – there is no way to program a hybrid platform. This involves being able to deploy operations on computers of different types (embedded, edge or cloud), using processing components of different types (multicores, GPUs, FPGAs etc) and all whilst resolving inter-node communications, synchronisation and optimisation problems.

This project will take a domain-specific approach to solving this problem. Starting from a program in a domain-specific application modelling package (such as TensorFlow), it will create a series of programming interfaces for each processor, device and the network, to allow applications to be rapidly and automatically deployed and optimised. It will demonstrate this capability on practical cyber-physical applications (such as in-vehicle Advanced Driver Awareness Systems) and on practical hybrid computing architectures.

The specific objectives of the project are:

  • Develop an understanding of the cyber-physical data analytics and AI applications and identify a promising candidate for further study.
  • Develop a toolset which allows an application developed in, for example, TensorFlow, to be realised on hybrid embedded/edge/cloud platforms.
  • Use the toolset to explore the design space for the chosen application, quantify the performance, cost and energy expenditure of various different implementations.
  • Demonstrator the ability to automatically derive efficient realisations automatically from the application model for your chosen use-case.
  • Present your work in leading international journals and conferences in the area. 



Contact details

Supervisor Name: John McAllister                                                                                        Tel: +44 (0)28 9097 1743

QUB Address: Institute of Electronics, Communications and Information Technology (ECIT)      Email: jp.mcallister@qub.ac.uk