Title: A Tightly Integrated System Processing Architecture for Next Generation Robotics
Project Type: iCASE
Imperial College London Supervisor: Prof. Paul Kelly / Dr. Christos Bouganis
Industry Partner: Dyson
Industry Supervisor: Dr. Rob Deaves
The next generation of domestic robots will mark a revolutionary change in capability. Current mass market, home based robots are restricted to vacuuming on a single floor level. The latest variants perform this role very effectively, but customer expectations are for much greater functionality. This will be enabled by the evolution of high fidelity 3-d sensing, anthropomorphic actuation suited for domestic environments, scene semantics provided by artificial intelligence, and ever massively-increased embedded processing power. Evidently, we have constituent components which should fuel game-changing domestic robots in the next few years.
An iCASE award has been made available by DYSON to support a research student in one of the following areas listed below; further support for strong candidates may be available from other funding sources:
Software Flexibility/Extensibility
Future robotics applications will depend on new developments in software tooling to enable complex software architectures to be developed, deployed, composed, combined, mapped, scheduled and ported efficiently. In robotics, complexity stems from having to write software for systems that are: distributed, heterogeneous, and energy constrained. Such platforms will combine heterogeneous resources, including both hardware accelerators – DSPs, FPGAs, GPGPUs, NPUs etc – and potentially also a mix of ARM, RISC-V, or x86 processors and microcontrollers. In this world it is beneficial to provide tools that allow algorithmic innovation whilst minimising the expense (or burden) of deploying new ideas onto distributed and heterogeneous hardware architectures.
An ideal candidate for this project should be excited about research spanning multiple layers of the robotic systems stack, and have experience in at least some of the following topics: compilers, operating systems, programming models for accelerators, software architectures for distributed, real-time or embedded systems, and appreciation of the challenges in robotics software.
HW Processing System
A critically important challenge in realising this potential, that forms the focus for this PhD project, is concerned with the higher level (on robot) processing requirements. The research will investigate distributed and heterogeneous hardware architectures capable to deliver the necessary performance including CPU, GPUs, Intelligent Processing Units (IPUs), dedicated hardware functions, reconfigurable elements, on-chip networks and memory systems. The outcome of the project will be insights to the architecture of System-on-Chip devices to support the next generation of domestic robots.
An ideal candidate for this project should be excited about research spanning multiple layers of the robotic systems stack, and have experience in at least some of the following topics: computer architecture (including accelerators), re-configurable microelectronics and robotic/image processing algorithms.
Information Flow
Another critically important area of the overall system, that requires further research effort is concerned with the integrated information architectures employed to efficiently gather ‘data’ and transform this to ‘information’ for control and decision making. The information will be distributed and exist in the cloud, on robot application processors and on robot real-time processors. Further, the information architecture will drive the requirements and research for databases, OS/RTOS, communication schemes, and processors and memory (SoCs).
An ideal candidate for this project should interested in creating the conceptual framework for ambitious integrated, resilient, robust, intelligent, reactive and safe robotics, and have experience in at least some of the following topics: operating systems, databases, distributed and event-based systems.
The collaboration between Imperial College London and Dyson will provide a unique opportunity for performing leading edge research that will have a direct conduit to next generation domestic robots!
If you are interested in applying please contact either Professor Paul Kelly or Dr Christos Bouganis for further details