University of North Texas, Texas, United States
Keywords: Edge Computing, Energy Efficiency, Real-Time Processing, Data Curation, Deep Neural NetworkThe traditional IoT has issues like high-energy consumption in cloud computing and communications, high communication latency due to transmission of lots of data from large of sensors to the cloud for any analytics. Edge datacenters (EDCs) are deployed to decrease the latency and networks congestion by processing data streams and user requests in near real-time. A large number of EDCs can be geographically distributed to offer mobile, low latency data transparency over real-time request and responses. In this project, we propose a new generation smart, secure edge datacenter (s-EDC) which can compress data before storing, which can adjust load, and with built-in deep neural network (DNN) can provide fast responses. We propose the following research for s-EDC: (1) Explore novel methods for real-time curation of sensor data at edge. (2) Explore novel load balancing techniques to authenticate the EDCs and find out less loaded EDC for task allocation. (3) Explore novel DNN architectures with multilayer active functions, novel training methods for efficient DNN training, and novel methods for accurate DNN training.