Bandwidth reductions gains through Edge Computing in connected cars
Abstract: Gathering and analyzing data that is generated in IoT and mobile devices is increasing due to the huge potential value it brings to consumers and car manufacturers. The increase in production of data introduces new problems in terms of bandwidth requirements. Performing computations, filtering, and analyzing data could be introduced to devices to reduce the bandwidth usage, this concept is called Edge computing. This Masters thesis has tackled the problem of bringing sensor based edge computing to vehicles. The problem motivation raised two questions; How arrival intensity of sensor data affects the system, and what bandwidth reductions gains can be made. Firstly a pipeline was defined, and technologies and frameworks were evaluated to be used in said pipeline in the method chapter. The PoC was then tested in a real car, in order to prove that it works. It was also used as a baseline for testing the two research questions posed. The PoC was then tested in two rounds, in order to evaluate the different research questions. The First research question was evaluated through having a static system and set of data and varying the sample rate of data to simulate the arrival intensity. The results show that the system had a linear relationship in terms of memory, cpu usage to the arrival rate. For the specific hardware that the system was tested showed that the system was stable up to a sample rate of 10000 Hz. The main research question was tested with the results from the secondary research question in mind. The sample rate was set to 100Hz and instead, the agent scenarios were varied in order to evaluate what bandwidth reductions gains can be made with three different edge computing levels. The results showed that the bandwidth can be reduced to 0.01% of the original amount when sampling data over 2 hours at 100 Hz. The scenarios had similar CPU usage despite increasing the amount of edge computing done in the agents, which further showed that edge computing is feasible in that car. However, it was also shown that the use case which the agent is based upon dictates what bandwidth reductions gains can be made.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)