Multi-User Simulation and Training for Forest Fire Management in Augmented Reality

The thesis work originates from an augmented reality application previously developed by the VR@POLITO laboratory and the Department of Automation and Computer Science at the Polytechnic University of Turin for training in forest dire management, used by the training centre of Corpo Volontari Anticendi Boschivi (AIB Corps). The source application, built in Unity, allowed learners to visualize and interact, on an individual basis, with a digital model of the environment and the forest fire under consideration. Tools were provided to dynamically curb and contain the expansion of the forest fire and it was possible to observe the consequences of one’s actions through integration with the open-source fire simulator Cell2Fire. The instructor projected the representation of the environment and fire expansion onto a sand table, making them visible to all, while individual learners, via a tablet, could interact with their digital model. The main purpose of this thesis was to design and provide a multi-user framework to the application, which would allow an instance of the wildfire to be managed by multiple learners simultaneously. In fact, the previous version of the application focused on providing a self-learning tool, going to foster a better understanding of fire evolution through a trial-and-error teaching approach. The learner could enter and remove one or more tactical interventions and view, with each confirmed change, the new fire front expansion produced by the simulator. The new version aims to take the application development one step further by retaining the previously implemented fire management logic but allowing multiple learners at the same time to take free action and later view the weighted result of the actions performed by individuals. The previous mode of use has been retained, and when creating the instance, one can choose whether to focus on the fire individually or to cooperate as a group. To achieve the aforementioned goal, two major changes were required: the Mirror library of Unity was introduced and used, providing both the appropriate data structures for real-time synchronization of user choices and the APIs to implement the client-server logic required for exchanging files during initialization and execution. The new version of the application is now capable of managing the tactical interventions of various learners at the server level instead of on individual client devices, avoiding duplication of files created by the simulator and reducing the information stored for each individual simulation. In addition, the new data management enabled the introduction of a new way of using the client-side application, implemented in a parallel thesis work, which allows learners to interact not only through a tablet but also with a visor (specifically, a Microsoft Hololens 2 device). This functionality takes advantage of pre-processing of files by the server, which relies on JSON data structures for storage and transfer between devices.