A Novel Spatio-Topological Embeddings for
Efficient & Model-free Redundant Node Placement
in 6G IoT Networks

Abstract: The development of 6G has accelerated the usage of IoT for data collection. With the continuous usage and ubiquitous connectivity, the nodes' batteries deplete soon and create the coverage hole in the network, imposing the connectivity challenges in 6G IoT. The problem has been dealt with a novel set of spatial features with topological embeddings extraction using graphical convolutional network. The deep deterministic policy gradient in the continuous action space trains the agent for the optimal placement of redundant nodes. The complete methodology with spatio-topological features have seen an improvement upto 13.11% in energy residual and 39.5% in the uniform load distribution than the state-of-the-art methods with stable network connectivity. Also the analysis at various environmental conditions with varying holes' density and sensors' density. The proposed scheme has shown the improvement under adverse coverage conditions too.

Keywords: 6G IoT, topological features, GCN, DDPG, Reinforcement Learning

1. Introduction

Moving towards a ubiquitous network with every device speaking with other, 6G era has gained tremendous growth. One important expectation in the 6G era is that machines and things will be the main consumers of mobile data traffic, and thus there will be more than 55 billion devices connected to the Internet by the end of 2025. These devices will continuously sense, process, act, and communicate with the surrounding environment, generating more than 70 zettabytes of data per year [1]. With the advancement of communication and AI, the 6G is supposed to handle massive communication, hyper-large AI data handling, reliable and low-latency communication, and ubiquitous connectivity [2]. The primary challenge and the basic requirement in this new era are continuous connectivity with devices like self driving cars, drones, IoT, robot operated factories etc. Traditional IoT applications have been advanced with the 6G communication capabilities which lead to more processing at IoT nodes using AI. The more the processing is done at local or edge devices or cloud, more is the battery consumption. The massive transmission in 6G imposes a challenge of continuous connectivity and longer battery life. This creates coverage holes in the network [3].

Many researchers have focused on the coverage hole issues. It involves the study into two phases: hole detection and hole recovery. The detection of holes is primarily focused on Voronoi diagram and Delaunay triangulation, which requires the exact geo-locations of the nodes, which is impractical and expensive for the large scale deployment...