Enabling peer-to-peer remote experimentation in distributed online remote laboratories

Enabling peer-to-peer remote experimentation in distributed online remote laboratories

Remote Access Laboratories (RALs) are online platforms that allow human user interaction with physical instruments over the Internet. Usually RALs follow a client-server paradigm. Dedicated providers create and maintain experiments and corresponding educational content. In contrast, this dissertation focuses on a Peer-to-Peer (P2P) service model for RALs where users are encouraged to host experiments at their location. This approach can be seen as an example of an Internet of Things (IoT) system. A set of smart devices work together providing a cyber-physical interface for users to run experiments remotely via the Internet.

The majority of traditional RAL learning activities focus on undergraduate education where hands-on experience such as building experiments, is not a major focus. In contrast this work is motivated by the need to improve Science, Technology, Engineering and Mathematics (STEM) education for school-aged children. Here physically constructing experiments forms a substantial part of the learning experience. In the proposed approach, experiments can be designed with relatively simple components such as LEGO Mindstorms or Arduinos. The user interface can be programed using SNAP!, a graphical programming tool.

While the motivation for the work is educational in nature, this thesis focuses on the technical details of experiment control in an opportunistic distributed environment. P2P RAL aims to enable any two random participants in the system - one in the role of maker creating and hosting an experiment and one in the role of learner using the experiment - to establish a communication session during which the learner runs the remote experiment through the Internet without requiring a centralized experiment or service provider. The makers need to have support to create the experiment according to a common web based programing interface. Thus, the P2P approach of RALs requires an architecture that provides a set of heterogeneous tools which can be used by makers to create a wide variety of experiments.

The core contribution of this dissertation is an automaton-based model (twin finite state automata) of the controller units and the controller interface of an experiment. This enables the creation of experiments based on a common platform, both in terms of software and hardware. This architecture enables further development of algorithms for evaluating and supporting the performance of users which is demonstrated through a number of algorithms. It can also ensure the safety of instruments with intelligent tools. The proposed network architecture for P2P RALs is designed to minimise latency to improve user satisfaction and learning experience. As experiment availability is limited for this approach of RALs, novel scheduling strategies are proposed.

Each of these contributions has been validated through either simulations, e.g. in case of network architecture and scheduling, or test-bed implementations, in case of the intelligent tools. Three example experiments are discussed along with users' feedback on their experience of creating an experiment and using others’ experimental setup. The focus of the thesis is mainly on the design and hosting of experiments and ensuring user accessibility to them. The main contributions of this thesis are in regards to machine learning and data mining techniques applied to IoT systems in order to realize the P2P RALs system.

This research has shown that a P2P architecture of RALs can provide a wide variety of experimental setups in a modular environment with high scalability. It can potentially enhance the user-learning experience while aiding the makers of experiments. It presents new aspects of learning analytics mechanisms to monitor and support users while running experiments, thus lending itself to further research. The proposed mathematical models are also applicable to other Internet of Things applications.