Software Architectures for Mobile Robots

Henrik Andreasson, Giorgio Grisetti, Todor Stoyanov, Alberto Pretto

Abstract

A software architecture defines the blueprints of a large computational system, and is thus a crucial part of the design and development effort. This task has been explored extensively in the context of mobile robots, resulting in a plethora of reference designs and implementations. As the software architecture defines the framework in which all components are implemented, it is naturally a very important aspect of a mobile robot system. In this chapter, we overview the requirements that the particular problem domain (a mobile robot system) imposes on the software framework. We discuss some of the current design solutions, provide a historical perspective on common frameworks, and outline directions for future development.

Introduction

Building a mobile robot is a massive integration challenge, requiring the coordination of high-speed sensor data, complex algorithms for SLAM and path planning, and precise hardware commands. Historically, researchers were forced to build custom "middleware"—the software glue connecting these disparate parts—from scratch, which often became a significant drain on development resources. Modern robotics has shifted toward standardized software architectures: high-level structures that determine how components communicate, scale, and remain robust under real-time constraints and asynchronous data processing.

Architectural Paradigms

The field has moved toward several key design patterns to manage complexity and ensure system reliability:

  • Component-Based Engineering: This approach breaks the robot's "brain" into independent nodes (e.g., driver nodes, planner nodes). This modularity allows for code reuse across different platforms and ensures that the failure of one node does not necessarily crash the entire system.
  • Communication Paradigms: Systems typically employ Publish-Subscribe models for continuous sensor streams and Client-Server models for discrete task requests.
  • Loose Coupling: Successful architectures prioritize loose coupling, where modules are self-sustained and interact through a standardized communication layer, such as an event bus.
An event communication bus connecting modular robotic components.

Comparison of Robotics Frameworks

The evolution of robotics middleware is marked by a transition from centralized discovery mechanisms to decentralized, peer-to-peer systems. While early versions of the Robot Operating System (ROS) relied on a central "master" node, newer iterations like ROS2 utilize Data Distribution Service (DDS) for automatic node discovery, enhancing reliability in distributed environments.

Comparison of major robotics software frameworks and their key features.

Various frameworks cater to specific niches within the community. For instance, YARP is frequently used in humanoid robotics, while Orocos is designed for hard real-time control. Despite these specialized tools, the industry has largely converged on hybrid solutions that combine peer-to-peer discovery with message-passing for data flow.

Key Insights and Future Directions

The transformation of middleware from a "necessary evil" into a standardized toolset has enabled unprecedented collaboration in the robotics community. However, several challenges remain for future development:

  1. Dynamic Load Balancing: Current frameworks struggle to automatically redistribute heavy computational tasks across a network based on available CPU resources.
  2. Bandwidth Management: Native support for dynamically adjusting data rates (e.g., reducing camera resolution during network congestion) is still lacking.
  3. Complexity Trade-offs: While modularity increases flexibility, it introduces communication overhead. Finding the optimal balance between a modular design and an integrated, high-performance system remains a primary challenge for developers.

As the field matures, the focus is shifting toward mapping specific robotic needs, such as fault tolerance, directly to architectural solutions like modular design and self-sustained modules.

Summary updated

Abstract

A software architecture defines the blueprints of a large computational system, and is thus a crucial part of the design and development effort. This task has been explored extensively in the context of mobile robots, resulting in a plethora of reference designs and implementations. As the software architecture defines the framework in which all components are implemented, it is naturally a very important aspect of a mobile robot system. In this chapter, we overview the requirements that the particular problem domain (a mobile robot system) imposes on the software framework. We discuss some of the current design solutions, provide a historical perspective on common frameworks, and outline directions for future development.
BibTeX
@misc{andreasson2022software,
  author = {Henrik Andreasson and Giorgio Grisetti and Todor Stoyanov and Alberto Pretto},
  title = {{Software Architectures for Mobile Robots}},
  year = {2022},
  eprint = {2206.03233},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2206.03233},
}
Figure