Electronic Thesis and Dissertation Repository

Thesis Format



Master of Engineering Science


Electrical and Computer Engineering

Collaborative Specialization

Artificial Intelligence


Katarina Grolinger


Rapid growth in numbers of connected devices, including sensors, mobile, wearable, and other Internet of Things (IoT) devices, is creating an explosion of data that are moving across the network. To carry out machine learning (ML), IoT data are typically transferred to the cloud or another centralized system for storage and processing; however, this causes latencies and increases network traffic. Edge computing has the potential to remedy those issues by moving computation closer to the network edge and data sources. On the other hand, edge computing is limited in terms of computational power and thus is not well suited for ML tasks. Consequently, this thesis aims to combine edge and cloud computing for IoT data analytics by taking advantage of edge nodes to reduce data transfer. Feature learning is performed on the edge with deep learning: the encoder part of the trained autoencoder is placed on the edge to extract relevant features. These features are sent to the cloud where the final ML task is performed directly, or the original features are first restored using the decoder part of the autoencoder. In the single-node model, all data are considered together and processed on a single edge node. In the multi-node model, sensors are grouped according to the IoT device locations or according to similarities between sensors, and different groups are processed on different edge nodes. The evaluation was performed on the task of human activity recognition from sensor data. Results show that data can be reduced on the edge up to 80% without significant loss in accuracy.

Summary for Lay Audience

Our daily lives are changing with rapid growth in numbers of smart and internet connected devices, including sensors, mobile, wearable, and other internet of things (IoT) devices. These devices are collecting data and transferring them to the cloud, possibly thousands of miles away, for storing and processing. Then, the results are sent back to the user for process control or decision making. However, transferring IoT data to the cloud is consuming network bandwidth, introducing latencies, and reducing the quality of service. Edge computing has the potential to remedy this situation by processing data close to where they are generated or collected. However, edge computing is not well suited for machine learning tasks as it has limited computation capabilities. Consequently, this thesis investigated merging edge and cloud computing for IoT analytics with the objective of reducing network traffic and latencies. The edge nodes apply machine learning to extract important features from the IoT data and send reduced data to the cloud for the final processing. The evaluation shows that for human activity recognition tasks, data can be reduced up to 80% without significant loss in accuracy.