Robot Motion Risk Reasoning Framework

TL;DR

This paper introduces a formal reasoning framework for robot motion risk, defining it with logic and probability, enabling explicit safety reasoning and risk-aware planning in complex environments.

Contribution

It provides the first formal definition of robot motion risk, unifies risk elements into a single metric, and reveals properties like non-additivity and history-dependency.

Findings

Formal risk definition using logic and probability.

Unified risk elements into a single metric.

Revealed properties like non-additivity and history-dependency.

Abstract

This paper presents a formal and comprehensive reasoning framework for robot motion risk, with a focus on locomotion in challenging unstructured or confined environments. Risk which locomoting robots face in physical spaces was not formally defined in the robotics literature. Safety or risk concerns were addressed in an ad hoc fashion, depending only on the specific application of interest. Without a formal definition, certain properties of risk were simply assumed but ill-supported, such as additivity or being Markovian. The only contributing adverse effect being considered is related with obstacles. This work proposes a formal definition of robot motion risk using propositional logic and probability theory. It presents a universe of risk elements within three major risk categories and unifies them into one single metric. True properties of risk are revealed with formal reasoning, such as non-additivity or history-dependency. Risk representation which encompasses risk effects from both temporal and spatial domain is presented. The resulted risk framework provides a formal approach to reason about robot motion risk. Safety of robot locomotion could be explicitly reasoned, quantified, and compared. It could be used for risk-aware planning and reasoning by both human and robotic agents.