Research Analyzer
← Back ICRA 2026

HEXAR: A Hierarchical Explainability Architecture for Robots

Tamlin Love, Ferran Gebellí, Pradip Pramanick, Antonio Andriella, Guillem Alenyà , Anaís Garrell, Raquel Ros, Silvia Rossi

PDF

AI summary

Key figure (auto-extracted from paper)
HEXAR significantly outperforms monolithic and aggregated baselines in accurately identifying root causes, excluding incorrect information, and reducing runtime for robotic explanations.
Explainable AI robotics hierarchical architecture modular explainability LLM reasoning autonomous systems

Problem

Existing robotic explainability approaches focus on isolated modules or use monolithic systems that ignore architectural modularity, making it difficult to generate accurate, high-level explanations for complex robots.

Approach

HEXAR is a plug-in hierarchical framework that routes user queries to specialized, module-tailored explainers via a selector, then aggregates their outputs into a single coherent explanation.

Key results

  • Higher root cause identification accuracy
  • Fewer incorrect facts in explanations
  • Reduced explanation generation runtime
  • Successful evaluation across 180 scenario-query variations on a TIAGo robot

Why it matters

Enables scalable, transparent decision-making for complex robotic systems, building trust and safety for end-users in assistive and autonomous applications.

Abstract

As robotic systems become increasingly complex, the need for explainable decision-making becomes critical. Ex- isting explainability approaches in robotics typically either focus on individual modules, which can be difficult to query from the perspective of high-level behaviour, or employ monolithic approaches, which do not exploit the modularity of robotic architectures. We present HEXAR (Hierarchical EXplainability Architecture for Robots), a novel framework that provides a plug-in, hierarchical approach to generate explanations about robotic systems. HEXAR consists of specialised component ex- plainers using diverse explanation techniques (e.g., LLM-based reasoning, causal models, feature importance, etc) tailored to specific robot modules, orchestrated by an explainer selector that chooses the most appropriate one for a given query. We implement and evaluate HEXAR on a TIAGo robot performing assistive tasks in a home environment, comparing it against end- to-end and aggregated baseline approaches across 180 scenario- query variations. We observe that HEXAR significantly outper- forms baselines in root cause identification, incorrect informa- tion exclusion, and runtime, offering a promising direction for transparent autonomous systems.

Index terms

Social HRI Natural Dialog for HRI Acceptability and Trust

Related papers