Robustification of Online Graph Exploration Methods

Exploring unknown environments is a fundamental task in many domains, e.g., robot navigation, network security, and internet search. We initiate the study of learning-augmented variant classical, notoriously hard online graph exploration problem by adding access to machine-learned predictions. propose an algorithm that naturally integrates predictions into well-known Nearest Neighbor (NN) significantly outperforms any known if prediction high accuracy while maintaining good guarantees when poor quality. provide theoretical worst-case bounds gracefully degrade with error, we complement them computational experiments confirm our results. Further, extend concept general framework robustify algorithms. By interpolating carefully between given NN, prove new performance leverage individual on particular inputs establishing robustness arbitrary inputs.