False positives and negatives in human-robot collision prevention: a virtual reality evaluation

Introduction: in human-robot collaboration (HRC), accurately detecting the proximity of the operator is essential to ensure safety without compromising productivity. However, conventional distance sensors face a trade-off between false positives (FP), which may cause unnecessary robot stops, and false negatives (FN), which can lead to dangerous collisions.Objectives: this study aims to analyze the balance between FP and FN in proximity detection within HRC environments using a virtual reality (VR) framework. The goal is to identify optimal sensor configurations that minimize FN without excessively increasing FP.Materials and Methods: a VR-based simulation environment was developed to evaluate different parameters: detection range, sensor beam angle, and sensor placement on the robot. An ANOVA model and post-hoc tests were applied to assess the statistical impact of each variable. Additionally, a second experiment was conducted involving a human participant to observe sensor behavior under realistic human intervention conditions.Results: the analysis identified sensor configurations that significantly reduce FN without notably increasing FP. Sensor beamangle and spatial coverage were found to be key factors. Tests involving human presence revealed additional challenges due tohuman movement variability and the need to fine-tune system sensitivity.Conclusions: the findings provide key technical criteria for sensor selection and configuration in collaborative applications. Strategies to mitigate FP are proposed, including the integration of advanced technologies such as computer vision and millimeter-wave radar sensors. This work contributes to the design of safer and more efficient systems for human-robot interaction in industrial settings.