University Thesis
↗Adaptive pathfinding algorithm for continuously learning agents in a traffic network, aimed at detecting road congestion. Agents improve their routing decisions over time by learning from the environment.
A traffic simulation where autonomous driver agents learn from experience and personalise their route choices - mimicking how real drivers navigate rather than just finding the shortest path.
Each driver has memory (remembering actual speeds on roads) and personality traits (stress tolerance, familiarity preference, learning rate) that shape their routing decisions. Roads dynamically slow down as traffic increases, creating realistic congestion that drivers learn to avoid or tolerate based on their individual preferences.
The system is evaluated against standard A* pathfinding and tested with Braess's Paradox scenarios to validate realistic traffic patterns.