Formation Control of UAV-fleet With Orientation Alignment: A Multi-Agent Actor-Critic Based Approach

Unmanned Aerial Vehicles (UAVs) have attracted much attention due to their application potential in many complex military and civilian operations, such as surveillance, event filming, and Search and Rescue (SAR) operations. Some of these operations often require the use of multiple UAVs to achieve higher efficiency while ensuring that specific mission requirements, such as resource management and maximization of covered areas, are met. In this work, we present a learning-based formation control protocol that pilots a networked fleet of UAVs configured in a leader-follower formation pattern to carry out a SAR mission. The formation consists of one leader, which is controlled by the Ground Control Station (GCS), and many followers that are expected to autonomously follow the leader while maintaining the original formation throughout the mission. The proposed protocol employs a Multi-Agent Reinforcement Learning (MARL) principle using only the Received Signal Strength Indicator (RSSI) derived during communication to autonomously control the follower UAVs to maintain formation while on a rotational movement. The goal is to ensure proper orientation alignment with the leader to avoid coverage overlap, and consequently increase efficiency. We performed several simulation experiments to evaluate the performance of the proposed protocol in terms of response time, and formation accuracy under different velocities.