Analyzing real-time communication in ROS2: a comparative study of DDS and Zenoh on autonomous drones

Robot Operating System 2 (ROS 2) is a widely adopted middleware framework designed to support modular, scalable, and real-time communication in robotic applications. Its architecture makes it particularly suitable for distributed systems, where multiple agents must coordinate over a network. In this context, the choice of ROS 2 middleware implementation (RMW) can significantly impact the performance of distributed robotic systems. This thesis investigates the communication behavior of the most widely adopted RMW implementations (Fast DDS, Cyclone DDS, Connext DDS, and Zenoh), to offer guidance in selecting the most appropriate one for an autonomous distributed drone system. The first objective of this thesis lies in developing a software suite to measure each RMW implementation’s performance in both synthetic and real traffic scenarios. The software suite is composed of three distinct benchmarks: one using middleware-specific APIs, while the others running on a ROS 2 infrastructure inside Docker containers. Moreover, we developed a real-time ROS 2 package to estimate and visualize communication metrics during real-world experiments, offering guidance for new design decisions, and help in troubleshooting problems. The results of the experiments show the tradeoff in terms of latency and reliability, and the problematic impact of congested networks on the communication between agents. This thesis outlines the need to design a communication infrastructure that exposes only inter-agent related traffic on the network, while keeping intraagent traffic local.