Abstract

Tactile Internet, as the next generation of the Internet, aims to transmit the modality of touch in addition to conventional audiovisual signals, thus transforming today’s content-delivery into skill-set delivery networks, which promises ultra-low latency and ultra reliability. Besides voice and data communication driving the design of the current Internet, Tactile Internet enables haptic communications by incorporating 5G networks and edge computing. A novel use-case of immersive, low-latency Tactile Internet applications is haptic-enabled Virtual Reality (VR), where an extremely low latency of less than 50 ms is required, which gives way to the so-called Remote Phobia treatment via VR. It is a greenfield in the telehealth domain with the goal of replicating normal therapy sessions with distant therapists and patients, thereby standing as a cost-efficient and time-saving solution.
In this thesis, we consider a recently proposed fog-based haptic-enabled VR system for remote treatment of animal phobia consisting of three main components: (1) therapist-side fog domain, (2) core network, and (3) patient-side fog domain. The patient and therapist domains are located in different fog domains, where their communication takes place through the core network. The therapist tries to cure the phobic patient remotely via a shared haptic virtual reality environment. However, certain haptic sensation messages associated with hand movements might not be reached in time, even in the most reliable networks. In this thesis, a prediction model is proposed to address the problem of excessive packet latency as well as packet loss, which may result in quality-of-experience (QoE) degradation. We aim to use machine learning to decouple the impact of excessive latency and extreme packet loss from the user experience perspective. For which, we propose a predictive framework called Edge Tactile Learner (ETL). Our proposed fog-based framework is responsible for predicting the zones touched by the therapist’s hand, then delivering it immediately to the patient-side fog domain if needed. The proposed ETL builds a model based on Weighted K-Nearest Neighbors (WKNN) to predict the zones touched by the therapist in a VR phobia treatment system. The simulation results indicate that our proposed predictive framework is instrumental in providing accurate and real-time haptic predictions to the patient-side fog domain. This increases patient’s immersion and synchronization between multiple senses such as audio, visual and haptic sensory, which leads to higher user Quality of Experience (QoE).