Using Behavior Trees in Risk Assessment

TL;DR

This paper introduces a novel approach using behavior trees to support early risk assessment in robotic missions, aiming to improve safety understanding during the design phase of cyber-physical systems.

Contribution

It presents the first use of behavior-tree models for risk assessment, facilitating early identification and visualization of risks in robotic system development.

Findings

Behavior-tree models aid early risk identification.

Practitioners find the approach improves understanding of risks.

The method bridges the gap between implementation and risk analysis.

Abstract

Cyber-physical production systems increasingly involve collaborative robotic missions, requiring more demand for robust and safe missions. Industries rely on risk assessments to identify potential failures and implement measures to mitigate their risks. Although it is recommended to conduct risk assessments early in the design of robotic missions, the state of practice in the industry is different. Safety experts often struggle to completely understand robotics missions at the early design stages of projects and to ensure that the output of risk assessments is adequately considered during implementation. This paper presents a design science study that conceived a model-based approach for early risk assessment in a development-centric way. Our approach supports risk assessment activities by using the behavior-tree model. We evaluated the approach together with five practitioners from four companies. Our findings highlight the potential of the behavior-tree model in supporting early identification, visualisation, and bridging the gap between code implementation and risk assessments' outputs. This approach is the first attempt to use the behavior-tree model to support risk assessment; thus, the findings highlight the need for further development.