Robotic systems increasingly operate as distributed, multi-host systems spanning robots, edge resources, and cloud services. In such settings, communication, localization, and resilience jointly determine system performance, yet existing research has often treated them separately. This thesis investigates how robotic capability can be strengthened in distributed ROS2 systems through the combined study of communication architecture, multisensor localization, and resilient edge deployment. The thesis first analyses communication in the edge–cloud continuum and shows, through a comparative study of CycloneDDS, MQTT, and Zenoh, that middleware performance is conditioned by deployment context: CycloneDDS is most effective in wired Ethernet environments, whereas Zenoh is better suited to Wi-Fi and 4G settings. The dissertation then addresses localization as a multisensor problem. It reviews event-based sensor fusion for odometry, investigates the effect of LiDARheterogeneity through comparative benchmarking of dome-shaped, solid-state, and spinning lidars, and evaluates seamless outdoor–indoor pedestrian positioning through GNSS/UWB/IMU fusion using error-state Kalman filtering, factor-graph optimization, and particle filtering. The findings show that localization accuracy, robustness, and continuity depend on the interaction between sensor characteristics, estimator structure, and environmental conditions; among the tested pedestrian positioning back-ends, the error-state Kalman filter yields the most consistent overall performance. Finally, the thesis examines resilience through a Kubernetes-orchestrated ROS2 multi-robot localization system for UWB-based relative positioning. Fault oriented experiments demonstrate that container orchestration can sustain localization quality through automated recovery even when edge-side services fail.