Reinforcement Learning for Decentralized Trajectory Design in Cellular UAV Networks with Sense-and-Send Protocol

TL;DR

This paper introduces a decentralized reinforcement learning approach for UAVs to optimize their trajectories in real-time sensing tasks over cellular networks, improving efficiency and data transmission success.

Contribution

It proposes a novel sense-and-send protocol and an enhanced multi-UAV Q-learning algorithm for decentralized trajectory design in dynamic environments.

Findings

Faster convergence of the proposed algorithm.

Higher utility achieved in real-time sensing scenarios.

Effective coordination among multiple UAVs.

Abstract

Recently, the unmanned aerial vehicles (UAVs) have been widely used in real-time sensing applications over cellular networks, which sense the conditions of the tasks and transmit the real-time sensory data to the base station (BS). The performance of a UAV is determined by the performance of both its sensing and transmission processes, which are influenced by the trajectory of the UAV. However, it is challenging for UAVs to design their trajectories efficiently, since they work in a dynamic environment. To tackle this challenge, in this paper, we adopt the reinforcement learning framework to solve the UAV trajectory design problem in a decentralized manner. To coordinate multiple UAVs performing the real-time sensing tasks, we first propose a sense-and-send protocol, and analyze the probability for successful valid data transmission using nested Markov chains. Then, we formulate the decentralized trajectory design problem and propose an enhanced multi-UAV Q-learning algorithm to solve this problem. Simulation results show that the proposed enhanced multi-UAV Q-learning algorithm converges faster and achieves higher utilities for the UAVs in the real-time task-sensing scenarios.