Content area
Robotics development is a challenging process that requires an understanding of how computers, robots, and the physical world interact. The success of this process depends on a programmer’s ability to mentally conceptualize the numerous potential states of a robotic system. Each additional variable, data stream, or environmental factor multiplies the possible states, leading to a combinatorial explosion of states that can quickly become unmanageable. To address this, the robotics community has developed data visualization tools that help programmers see system states rather than reconstruct them mentally. Although these tools are now integral to the development workflow, visualization tools have not kept pace with the rapid advances in robotics systems. As a result, the development process remains complex and error-prone. This dissertation explores how to design effective robotics data visualization tools, how to integrate them seamlessly into the development process, and how they can improve our ability to build robotics systems. I introduce a visualization-driven robotics development (VDRD) approach, which advocates for tightly integrating data visualizations into the development workflow. I examine this concept through four studies: the first two investigate the impact of visualization tools on traditional development practices, while the latter two evaluate how applying synthesized design guidelines can enhance VDRD in both current and emerging robotics workflows. Ultimately, this research provides a foundation for building tools that augment our ability to understand and develop robotics systems more effectively.