Metaverse in InterPlanet Internet: Reinforcement Learning Implementation of Time-Dependent Machine Learning Model to Make Robots Fit for Space Applications

EasyChair Preprint 9532

Abstract

The interplanet internet is a conceived computer network in space, consisting of a set of network nodes that can communicate with each other. These nodes are the planet’s orbiters (satellites) and landers (e.g. robots, autonomous machines, etc.) and the earth ground stations. As resource depletion on Earth becomes real, the idea of extracting valuable elements from asteroids or using space-based resources to build space habitats becomes more attractive, one of the key technologies for harvesting resources is robotic space mining or robotic building of space settlement. The metaverse is essentially a simulated digital environment mimicking the real world. The metaverse would be something very similar to real world planetary activities where users( space colonies or internet users on Earth) interact with overlaying objects represented by robots, drones, etc. for real-world planetary activities like space mining, etc. In this paper, we use information about different time steps as represented by the observation matrix for the presence of the robots in a planetary environment by  encoding the robots presence as reinforcement learning agent on the site at different time steps. The reinforcement learning agent uses deep convolutional neural networks with Q-learning algorithm to approximate the Q-function and that uses experience replay. The neural network used by the learning agent is a time dependent encoded AI model and trained with Reinforcement learning by different methods to make an autonomous robots to learn. The results of the study simulated on existing  internet here on Earth show that the real individual  behaviour on a distant planet can be achieved  provided the interplanet internet is available as pathway communication. Therefore, connected metaverse with  time-dependent encoded layers of virtual spaces along with deep learning models with learning agents could be of reality even in interplanet environment.

Keyphrases: Deep Q-Learning, InterPlanet Internet, Metaverse, Reddy's Encoding Model, Reinforcement Learning

@booklet{EasyChair:9532,
  author    = {Poondru Prithvinath Reddy},
  title     = {Metaverse in InterPlanet Internet: Reinforcement Learning Implementation of Time-Dependent Machine Learning Model to Make Robots Fit for Space Applications},
  howpublished = {EasyChair Preprint 9532},
  year      = {EasyChair, 2023}}