Intelligent Few-shot based Real-time Traffic planning system for autonomous Vehicles using Graph SAGE-LSTM with transfer learning

This research presents a detailed framework for predicting traffic speed by merging spatial and temporal aspects using an advanced hybrid architecture of GraphSAGE and Long Short-Term Memory (LSTM) networks, enhanced through transfer learning methodologies. Central to this approach is the representation of traffic speed data collected from loop detectors as a graph structure, wherein each sensor acts as a node. The GraphSAGE convolutions extract complex spatial relationships among the nodes, allowing the model to learn how traffic speeds influence each other based on geographic proximity and sensor interactions. To capture the dynamic temporal dependencies inherent in traffic behaviours, the LSTM networks are employed, which excel at recognising patterns over time by maintaining information across long sequences. This architecture is particularly beneficial for predicting traffic speed fluctuations based on historical data trends. The framework begins with a pretraining phase on the PEMS-BAY dataset, which comprises extensive traffic speed records. Subsequently, the model undergoes fine-tuning on the METR-LA dataset, utilising a limited number of training batches. Limited-batch fine-tuning based on transfer learning is a solution for adapting to data-scarce conditions under the few-shot learning framework that is independent of traditional episodic meta-learning formulations, which require multi-task support-query sets. It is based on the conditions encountered in practice, where the labeled data for the new traffic scenarios is normally very limited. This simulates a few-shot learning scenario, reflecting real-world conditions where labelled data for new traffic environments is often minimal. Through extensive experimentation, the results demonstrate that integrating spatial graph learning via GraphSAGE, temporal modelling via LSTM, and transfer-based limited-batch fine-tuning used here as a practical surrogate for episodic few-shot learning significantly improves traffic speed prediction accuracy, particularly in data-scarce environments. The proposed framework achieves strong performance on the METR-LA benchmark while using only 20 training batches for adaptation, validating its applicability in intelligent traffic management for autonomous vehicles.