Towards Declarative Safety Rules for Perception Specification Architectures

TL;DR

This paper analyzes vision pipelines in agricultural robots to develop declarative safety rules, enabling verification of system operation and enhancing safety standards for autonomous agricultural machinery.

Contribution

It introduces a method for specifying safety rules declaratively and verifying vision pipeline functionality to improve safety compliance in agricultural robots.

Findings

Proposed a verification mechanism for vision pipelines.

Enhanced safety assurance in autonomous agricultural systems.

Demonstrated applicability of safety rules in real-world scenarios.

Abstract

Agriculture has a high number of fatalities compared to other blue collar fields, additionally population decreasing in rural areas is resulting in decreased work force. These issues have resulted in increased focus on improving efficiency of and introducing autonomy in agriculture. Field robots are an increasingly promising branch of robotics targeted at full automation in agriculture. The safety aspect however is rely addressed in connection with safety standards, which limits the real-world applicability. In this paper we present an analysis of a vision pipeline in connection with functional-safety standards, in order to propose solutions for how to ascertain that the system operates as required. Based on the analysis we demonstrate a simple mechanism for verifying that a vision pipeline is functioning correctly, thus improving the safety in the overall system.