Overview
Machine learning and deep learning techniques have become the standard for traffic prediction. Most significant applications, such as those offered by Google and Apple, rely heavily on historical data. However, accurately factoring in the impact of sudden events—like accidents or wrong-way drivers—remains a challenge, mainly when delivering real-time updates to drivers and transportation companies. Another limitation is the lack of transparency in travel time predictions, making it difficult for users to understand or adjust routes accordingly.
For logistics companies, estimating the effects of traffic congestion on their supply chain and understanding the causes have become crucial to keeping their operational costs low and ensuring the ever-growing demand for just-in-time delivery. Furthermore, most route planning applications in the market are targeted at common drivers rather than the heavy transportation sector.
This project combines science-based innovation with business-driven research to create a highly adaptive and explainable traffic prediction model by applying state-of-the-art methods and including causal machine learning (CML) and graph neuronal networks (GNN). This model will account for the effects of unforeseen events when calculating travel times between locations. Initially, it will be tailored to help logistics companies meet delivery deadlines and optimize their supply chains.
