Identifying how a digital entity could best represent a physical entity within a heavily constrained network in the operational stage of a Digital twi
Lead Research Organisation:
Newcastle University
Department Name: Sch of Computing
Abstract
My thesis aims to helps businesses that employ digital twins who want to improve a physical asset's performance by identifying the impact of data degradation upon the connection between the physical and digital entity and creating solutions to rectify any performance reduction.
This project's initial aim is to develop an environment that would enable further analysis of data acquisition and processing problems within digital twin. This environment involves simulating a digital twin and understanding how data degradation affects communication between a physical and a digital entity. Once ample knowledge is derived from this, an appropriate method of rectifying poor performance due to data degradation will be developed.
The physical entity in this environment will be a car object within a simulator. The ease, simplicity, and speed at which development would occur using a simulated car for data degradation experimentation would drastically outweigh the time taken to research, develop, and construct a physical car model. Much thought has been conducted into the simulator for development, specifically into Microsoft AirSim, and CARLA open-source simulator. From the initial outlook, the CARLA simulator is preferred for this project. The simulator was developed to support the development, training, and validation of the autonomous system. This simulator has a suite of sensors, environmental conditions, and dynamic actors to simulate a digital twin.
This project's digital entity will be an AI controller developed with reinforcement learning, which will receive data from sensors associated with the simulated environment's vehicle model. The controller will also provide commands to the vehicle entity in the simulated environment. The relationship between the physical vehicle and the AI controller will be a closed-loop system without human interference.
The policies that govern data degradation will encompass three separate approaches: a naive approach, a data compression approach, and a data-driven approach. The naïve approach will involve simplistic policies such as introducing latency into the physical and digital entities' communication or arbitrarily reducing the amount of information transmitted between the two entities. The data compression approach will involve utilizing gallium data compression algorithm to identify the relationship between the compression percentage and the physical entity's performance. Finally, the data-driven approach aims to identify data that has the highest impact on the physical entity's performance.
Once the separate components are developed, an evaluation of the performance will be conducted, which will involve the physical car driving in a circuit, measure the time taken between checkpoints, and with different data degradation policies to ascertain the impact on the performance of the physical entity.
Upon completion of the evaluation regarding the impact the data degradation has on the digital twin, ample research a will be conducted into rectifying the impact on the data degradation.
This project's initial aim is to develop an environment that would enable further analysis of data acquisition and processing problems within digital twin. This environment involves simulating a digital twin and understanding how data degradation affects communication between a physical and a digital entity. Once ample knowledge is derived from this, an appropriate method of rectifying poor performance due to data degradation will be developed.
The physical entity in this environment will be a car object within a simulator. The ease, simplicity, and speed at which development would occur using a simulated car for data degradation experimentation would drastically outweigh the time taken to research, develop, and construct a physical car model. Much thought has been conducted into the simulator for development, specifically into Microsoft AirSim, and CARLA open-source simulator. From the initial outlook, the CARLA simulator is preferred for this project. The simulator was developed to support the development, training, and validation of the autonomous system. This simulator has a suite of sensors, environmental conditions, and dynamic actors to simulate a digital twin.
This project's digital entity will be an AI controller developed with reinforcement learning, which will receive data from sensors associated with the simulated environment's vehicle model. The controller will also provide commands to the vehicle entity in the simulated environment. The relationship between the physical vehicle and the AI controller will be a closed-loop system without human interference.
The policies that govern data degradation will encompass three separate approaches: a naive approach, a data compression approach, and a data-driven approach. The naïve approach will involve simplistic policies such as introducing latency into the physical and digital entities' communication or arbitrarily reducing the amount of information transmitted between the two entities. The data compression approach will involve utilizing gallium data compression algorithm to identify the relationship between the compression percentage and the physical entity's performance. Finally, the data-driven approach aims to identify data that has the highest impact on the physical entity's performance.
Once the separate components are developed, an evaluation of the performance will be conducted, which will involve the physical car driving in a circuit, measure the time taken between checkpoints, and with different data degradation policies to ascertain the impact on the performance of the physical entity.
Upon completion of the evaluation regarding the impact the data degradation has on the digital twin, ample research a will be conducted into rectifying the impact on the data degradation.
Organisations
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
EP/R51309X/1 | 01/10/2018 | 30/09/2023 | |||
2448711 | Studentship | EP/R51309X/1 | 01/10/2020 | 30/09/2023 | Tolulope Awosanya |