AI-Driven Adaptive Autonomy: Is AI Really Pervasive? Research Gaps from Bibliometric Assessment
Simona Casini, Andrea Caiti, Pietro Ducange, Francesco Marcelloni, Lorenzo Pollini
AI summary
Problem
Existing bibliometric studies map research themes but fail to show how AI contributions distribute across the functional architecture of autonomous systems. This paper asks whether AI is truly pervasive across the core functions enabling adaptive autonomy and identifies which areas are mature versus under-explored.
Approach
The authors adapted the MAPE-K control-loop framework into a 13-module functional architecture and mapped over 2,500 AI-robotics publications using a multi-label neural classifier. Co-occurrence and structural network analyses then revealed where AI integration is concentrated or lacking.
Key results
- Fine-grained 13-module functional architecture mapping AI across core robotic functions
- Reproducible multi-label neural classifier achieving >90% F1-score for literature mapping
- Identification of mature AI integration areas alongside surprising gaps in diagnostics and knowledge management
- Structural co-occurrence analysis revealing strong perception-planning links but weak strategy-adaptation connections
Why it matters
Provides roboticists and AI researchers with an architectural lens to identify underexplored functional gaps and guide future priorities in human-centered adaptive autonomy.
Abstract
Artificial Intelligence is widely recognised as a driver of adaptive autonomy in robotics. Yet, the extent to which AI techniques truly permeate the functional architecture of au- tonomous systems is still only partially characterised. Existing bibliometric analyses typically map research themes,