Optimizing Inter-Process Communication for Robotics Applications

Robots have become increasingly complex and computationally capable, resulting in equally complex software architectures composed of numerous modules that must exchange large volumes of data. ROS and its successor, ROS 2, have become the de facto standards for managing inter-process communication (IPC) in robotics. However, in its default configuration, ROS 2 suffers from significant performance degradation when communicating over unreliable networks, negatively impacting both local and remote communication. Additionally, its overhead grows substantially with the number of nodes and communication interfaces. This thesis proposes an alternative IPC solution that is fast, reliable, deterministic, efficient, and easy to use, with a strong focus on robust local communication. The solution comprises a library for local IPC and a gateway application for extending communication across networks. Experimental results demonstrate that the proposed approach effectively mitigates the performance issues of ROS 2 under unreliable network conditions. Compared to alternative ROS 2 middleware configurations addressing similar challenges, the solution consistently delivers lower latency, reduced CPU and memory usage, and improved scalability.