Visualizing Spatial-temporal Data in Robotics: An Overview

Robots continuously generate massive amounts of spatial–temporal data as they perceive, predict, and act within dynamic environments. Although many visualization frameworks have been developed in other fields, these methods cannot be directly applied to robotics. Robotic data streams are high-frequency, uncertainty-laden, and tightly coupled with real-time decision-making, which introduces challenges not present in traditional geospatial settings. This literature review synthesizes foundational models for spatial–temporal data visualization, including the Triad Framework, Visual Analytics Framework, and methods such as space–time cubes, heat maps and flow maps by evaluating their applicability to robotic contexts. We identify three core challenges that lead to the gap in visualizing spatial-temporal data for robotics: uncertainty, real-time constraints and human cognitive load. With insights from prior work in visualization and human–robot interaction, we outline a set of design principles for developing visualization frameworks tailored specifically to robotics. To demonstrate these principles concretely, we present an interactive prototype that compares three visualization conditions, including linked overall confidence, linked subtask confidence, and integrated subtask confidence, designed to support operator supervision of a search and rescue robot. These principles and the accompanying prototype aim to support interpretability, reduce cognitive burden, and ultimately enable more effective human–robot collaboration. A pilot study with six participants suggests that the linked subtask view offers the most reliable detection of robot state changes, while the integrated view supports the most accurate diagnosis but is more visually demanding.