Solving Tactile Internet By Faking Physics

Abstract

Tactile Internet (TI) allows kinesthetic interactions with a remote environment and haptic sensory feedback over a network, essentially adding a new sensory dimension to the internet. TI has a wide range of applications such as enabling remote work for professions that require human hands and tactile sensory input like car repairs or medical operations. Most TI applications have ultra low latency requirements that limit TI to short distances because the speed of light poses a constraint on how fast data can be transmitted over a distance. This thesis attempts to relax the delay requirements for enabling TI over long distances. This is achieved by locally simulating the physical interactions and tactile properties of the environment to provide instant haptic feedback regardless of network conditions.

Model Mediated Teleoperation (MMT) creates a local feedback loop using a model and lowers the network delay requirements of the system at large. We propose utilizing a physics engine with haptic functionalities as the local model, to create MMT based TI applications that can cater for a variety of complex scenarios. We implement a multi-purpose haptic framework attached to Unity that facilitates creating various TI testbeds and applications, and use it to analyze and benchmark our MMT design. We show that haptic feedback from stationary objects is completely unaffected by network delay with our MMT implementation, and narrow down the challenge to simulating dynamic objects. We demonstrate that the model inevitably diverges from the environment it simulates when movable objects are involved, and needs frequent corrections.