As collaborative robots (cobots) increasingly share workspaces with humans, ensuring safe physical human–robot interaction (pHRI) has become paramount. This systematic review addresses safety assessment in pHRI, focussing on the industrial field, with the objective of collecting approaches and practices developed so far for modelling, simulating, and verifying possible collisions in human–robot collaboration (HRC). To this aim, advances in human–robot collision modelling and test-based safety evaluation over the last fifteen years were examined, identifying six main categories: human body modelling, robot modelling, collision modelling, determining safe limits, approaches for evaluating human–robot contact, and biofidelic sensor development. Despite the reported advancements, several persistent challenges were identified, including the over-reliance on simplified quasi-static models, insufficient exploration of transient contact dynamics, and a lack of inclusivity in demographic data for establishing safety thresholds. This analysis also underscores the limitations of the biofidelic sensors currently used and the need for standardised validation protocols for the impact scenarios identified through risk assessment. By providing a comprehensive overview of the topic, this review aims to inspire researchers to address underexplored areas and foster innovation in developing advanced, but suitable, models to simulate human–robot contact and technologies and methodologies for reliable and user-friendly safety validation approaches. Further deepening those topics, even combined with each other, will bring about the twofold effect of easing the implementation while increasing the safety of robotic applications characterised by pHRI.

Ensuring the safety of physical human–robot interaction (pHRI) is of utmost importance for industries and organisations seeking to incorporate robots into their workspaces. To address this concern, the ISO/TS 15066:2016 outlines hazard analysis and preventive measures for ensuring safety in Human–Robot Collaboration (HRC). To analyse human–robot contact, it is common practice to separately evaluate the “transient” and “quasi-static” contact phases. Accurately measuring transient forces during close human–robot collaboration requires so-called “biofidelic” sensors that closely mimic human tissue properties, featuring adequate bandwidth and balanced damping. The dynamics of physical human–robot interactions using biofidelic measuring devices are being explored in this research. In this paper, one biofidelic sensor is tested to analyse its dynamic characteristics and identify the main factors influencing its performance and its practical applications for testing. To this aim, sensor parameters, such as natural frequency and damping coefficient, are estimated by utilising a custom physical pendulum setup to impact the sensor. Mathematical models developed to characterise the sensor system and pendulum dynamics are also disclosed.

The integration of collaborative robots (Cobots) into various fields is increasingly observed, and the importance of ensuring safety in human-robot interactions (HRI) is thereby underscored. The critical aspect of transient contact dynamics, a fundamental yet under-explored area in HRI safety, is addressed in this paper. Unlike quasi-static contacts, transient interactions involve brief, dynamic impacts that pose unique challenges for safety assessments and Cobot behaviour modulation. In this paper, a novel approach to studying these interactions is introduced, employing a customizable pendulum experimental setup that simulates transient contact scenarios. Additionally, a comprehensive model for calculating the equivalent mass of robots is developed, aiming at improving the simplified model currently proposed by ISO/TS 15066. The robot’s configuration and the direction of impact are incorporated into this model, providing a more nuanced understanding of the forces at play during contact. A series of experiments demonstrates the relationship between impact forces, robot velocity, and the simulated mass of human body parts. These insights are deemed useful for the design of Cobots to minimize the risk of injury in a shared workspace.