Interfacing ns-3 and ROS2 for Edge-Controlled Robotic Systems: Impact of Network Dynamics on Performance

Modern robotic systems increasingly rely on edge computing to offload computationally intensive tasks while maintaining real-time responsiveness. In such distributed architectures, the quality of the underlying communication network becomes a decisive factor for control stability, safety, and overall system performance. Latency, packet loss, and other impairments can significantly degrade closed-loop behavior, yet these factors are often overlooked in conventional robotic simulation environments that assume ideal network conditions. This thesis addresses this gap by integrating the Robot Operating System 2 (ROS2) with the ns-3 network simulator, thereby creating a co-simulation framework capable of modeling realistic network dynamics in edge-controlled robotic systems. The proposed methodology combines Gazebo-based physical simulation, Linux network namespaces to emulate distributed environments, and ns-3 as a configurable network bridge between robot and controller. Tools such as iperf3 and ping were initially employed to validate network impairments, followed by detailed experiments on robotic control performance using ROS2 topics under varying conditions of latency and packet loss. Through systematic experiments, the work demonstrates how even moderate latencies can accumulate to noticeable control delays, while packet loss, although often masked under constant velocity commands, becomes significant under more dynamic inputs. By employing hybrid configurations, including Linux tc netem to complement ns-3 propagation models, the framework overcomes challenges in real-time scheduling and provides a more faithful emulation of wide-area delays. Results analyzed with PlotJuggler reveal the interplay between commanded and measured robot velocities across scenarios ranging from perfect networks to combined high-latency and lossy channels.