Hardware Development for Cyber-Physical Systems adaptive control algorithms

Lead Research Organisation: University of Southampton
Department Name: Electronics and Computer Science

Abstract

Cyber-Physical Systems (CPS) are integrations of computation with physical processes.
Embedded computers and networks monitor and control the physical processes, usually with feedback loops where physical process affect computations and vice versa. CPS gained a lot of importance from last decade because of its self-adaptive control nature.
To this end, there are several challenges need to be addressed for providing accurate self adaptive control to CPS, including physical model identification in real time when the system subjected to external disturbances and in the presence of uncertainty. Recently proposed estimation based multiple model switched adaptive control (EMMSAC) are shown that the best performance in the presence of uncertainty and external disturbances over the state-of-the-art algorithms for controlling physical system adaptively.
These algorithms have been implanting on the Digital Signal Processing (DSP) platform. In general, these algorithms need a lot of computing resources and reduced execution time. To cope with all these challenges, designers can rely on more and more mature digital electronics technologies. Further state-of-the-art implementations are limited by speed. Moreover in cyber-physical system if we desired to control physical system adaptively, EMMSAC should perform operations in micro or nanoseconds thereby physical system will able to adapt the controller instructions within the scheduled time. To achieve aforementioned characteristics we are intended to develop these EMMSAC algorithms on field programmable gate array (FPGA). This will help to build an efficient, stable, accurate and reliable cyber-physical system. In the initial stage of work, we developed a 2x2 kalman filter on Virtex ultra-scale FPGA, which is the fundamental block in the EMMSAC algorithm. We have also developed a case study for controlling the velocity of the car by adapting the road conditions.
In future, we are intended to apply these algorithms for boost converter which we can develop in our lab environment under different load conditions. Also we are intended to increase number models 50-400. The next future challenge is to implement the dynamic
online refinement strategy for EMMSAC.

Publications

10 25 50
publication icon
Vala C (2019) Low-Complexity Architecture for Cyber-Physical Systems Model Identification in IEEE Transactions on Circuits and Systems II: Express Briefs

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509747/1 01/10/2016 30/09/2021
1922711 Studentship EP/N509747/1 01/03/2017 31/08/2020 Charan Vala
 
Description We proposed a low complexity architecture for cyber-physical system (CPS) model identification based on multiple-model adaptive estimation (MMAE) algorithms. The complexity reduction is achieved by reducing the number of multiplications
in the filter banks of the MMAE algorithm present in the cyber
component of the CPS. The architecture has been implemented
using FPGA for 16, 32, 64 filter banks as part of position and
velocity estimations of autonomous auto-mobile application. It
has been found up to 78% reduction in multiplications is possible, which translates to the reduction of 39% lookup tables,
13% FFs, 27% DSPs, and 43% power reduction when compared with the conventional architecture (without multiplications
reduction) at 100MHz operating frequency. Furthermore, the
proposed architecture is able to identify the accurate model of automobile application just within 510 ns, in the presence of external
disturbances and abrupt changes.
Exploitation Route Recent research is focusing on the Cyber-physical
system (CPS) Model-identification (MI) due to the
increasing demand for applications such as system identification intelligent control system, reinforcement learning, system faults detection and isolation, enhanced
closed-loop control systems aim to find optimal decisions
(decision rules) in unknown or uncertain environments using
a qualitative and noisy on-line performance feedback. Generally, the physical system is classified as linear time-invariant (LTI) and variant (LTV). It is also worth noting that dynamical time-invariant system is a special case of dynamical time-variant systems, i.e., dynamical time-variant systems apparently have more general appeal, because most industrial
dynamical systems are inherently time-variant. For example,
the parameters of components in an electrical circuit system
often vary with time. On the other hand, in many applications
such as aerospace, process control etc., the ever-increasing
performance demands and more stringent specifications over a
wide range of operating conditions diminish the value of time-
invariant models as good approximations of the actual plant.
Time-variant systems, however, exhibit more complicated dynamics, therefore the model identification of LTI and LTV systems is challenging. Recently based
on Multiple models adaptive estimation (MMAE) algorithm which works based on Kalman filter bank, we proposed LTI and LTV model identification hardware architecture and validated on Virtex ultra-scale + FPGA.
Sectors Aerospace, Defence and Marine,Electronics,Energy