Dynamic Risk Assessment for Human-Robot Collaboration Using a Heuristics-based Approach

TL;DR

This paper tackles the safety challenges of Human-Robot Collaboration by proposing a Dynamic Risk Assessment framework that generates objective hazard indicators from scene parameters and a new anthropocentric head-orientation measure (PHH). Hazard indicators for distance, velocity (including direction), and PHH are non-linearly mapped to the unit interval and weighted to form a Total Hazard Indicator $R_{total}=\frac{1}{\sum_i\omega_i}\sum_i\omega_i r_i$, with $\omega_1=\omega_2=1$ and $\omega_3=2$, ensuring a transparent, data-light risk valuation. The PHH indicator captures human awareness and its deviations from a reference viewpoint toward the robot, which, combined with distance and velocity cues, can reveal hazardous conditions that may not be apparent from motion alone. The ABiD case study demonstrates that identical robot motions can yield different hazard profiles depending on human behavior, underscoring the value of a holistic, indicator-based hazard analysis for dynamic risk assessment in real industrial settings.

Abstract

Human-robot collaboration (HRC) introduces significant safety challenges, particularly in protecting human operators working alongside collaborative robots (cobots). While current ISO standards emphasize risk assessment and hazard identification, these procedures are often insufficient for addressing the complexity of HRC environments, which involve numerous design factors and dynamic interactions. This publication presents a method for objective hazard analysis to support Dynamic Risk Assessment, extending beyond reliance on expert knowledge. The approach monitors scene parameters, such as the distance between human body parts and the cobot, as well as the cobot`s Cartesian velocity. Additionally, an anthropocentric parameter focusing on the orientation of the human head within the collaborative workspace is introduced. These parameters are transformed into hazard indicators using non-linear heuristic functions. The hazard indicators are then aggregated to estimate the total hazard level of a given scenario. The proposed method is evaluated using an industrial dataset that depicts various interactions between a human operator and a cobot.

Figures (8)

• Figure 1: Distance-based hazard indicator for different values of $\alpha$ with $d_\mathrm{min} = 0$.
• Figure 2: Heat map of the velocity-based hazard indicator.
• Figure 3: Plot of the PHH-based hazard indicator for different values of $c$.
• Figure 4: Example taken from the ABiD Dataset showcasing a handover scenario. © Proximity Robotics & Automation GmbH
• Figure 5: Comparison of the total hazard for the handover scenarios.