Prominence-guided link prediction in fuzzy robotic networks

Area of Science:

• Cyber-Physical Systems (CPSs)
• Robotics and Intelligent Automation
• Network Science

Background:

• Increasing complexity in cyber-physical systems (CPSs) necessitates robust predictive models for dependable operations.
• Existing fuzzy link prediction methods face challenges in accurately modeling uncertain interactions within robotic networks.
• The Strength Prominence (SP) index, originally for social networks, requires adaptation for CPS applications.

Purpose of the Study:

• To reformulate the Strength Prominence (SP) index for fuzzy interaction graphs in robotic and intelligent automation systems.
• To enhance the prediction of link probabilities in cyber-physical systems (CPSs) by considering connectedness and prominence.
• To validate the modified SP index's performance against traditional fuzzy indices for link prediction.

Main Methods:

• Reformulation of the Strength Prominence (SP) index for fuzzy interaction graphs, where nodes represent robotic components and edges denote uncertain dependencies.
• Assessment of link probability using connectedness strength and prominence levels, even without common neighbors.
• Theoretical demonstration of properties like symmetry, boundedness, and monotonicity.

Main Results:

• The modified SP index demonstrated superior predictive accuracy compared to traditional fuzzy indices (CN, RSM, CAR) on real-world and ROS-based robotic datasets.
• Achieved higher precision, AUC (Area Under the Curve), and AUP (Area Under the Precision-Recall Curve) measurements.
• Successfully identified interaction failures early and improved collaboration prediction.

Conclusions:

• The adapted SP index offers a novel interdisciplinary tool for fuzzy link prediction in cyber-physical systems (CPSs).
• This approach aids in the development of fault-tolerant designs and resilient network structures for autonomous systems.
• Enables improved real-time robotic collaboration and enhances the dependability of complex robotic networks.