Multimodal Ai for Smart Transportation Systems

Multimodal AI for Smart Transportation Systems is an interdisciplinary international workshop that brings together the power of multimodal machine learning with urban mobility challenges. Participants will explore how to integrate data from traffic cameras, LIDAR, GPS, smart infrastructure, and language-based inputs (e.g., reports, driver commands) to enable AI-based decision-making in urban and intercity transport networks.

Through practical case studies and hands-on labs, participants will learn to build models for traffic flow prediction, incident detection, vehicle identification, smart routing, and policy compliance using tools such as PyTorch, Hugging Face Transformers, OpenCV, DeepStream, and Spatio-temporal Graph Networks.

• Introduce participants to multimodal AI principles and toolkits

• Enable fusion of heterogeneous data sources for decision-making

• Promote innovation in congestion reduction, public safety, and autonomous navigation

• Demonstrate real-world smart transport use cases through hands-on labs

• Foster collaboration between AI technologists and transport professionals

📅 Day 1 – Fusing Traffic CCTV, GPS, and V2X Streams

• Introduction to multimodal data in intelligent transportation systems (ITS)

• Overview of traffic data sources: CCTV footage, GPS traces, and V2X (vehicle-to-everything) communication

• Data fusion techniques: early, late, and hybrid fusion approaches

• Synchronization and preprocessing of heterogeneous data streams

• Object detection and tracking from CCTV using deep learning

• GPS-based trajectory extraction and analysis

• V2X communication data: protocols and use cases

• Real-time data pipelines for multimodal integration

📅 Day 2 – Spatio-Temporal Graph Networks for Congestion Prediction

• Understanding traffic as a dynamic spatio-temporal graph

• Basics of Graph Neural Networks (GNNs)

• Temporal dynamics with Recurrent and Transformer models

• Building spatio-temporal graph neural networks (ST-GNNs)

• Feature engineering for nodes (intersections) and edges (roads)

• Modeling congestion patterns and hotspot detection

• Dataset sources: METR-LA, PeMS, OpenTraffic

• Evaluation metrics: MAE, RMSE, MAPE for prediction models

📅 Day 3 – Decision Support for Adaptive Signalling

• Introduction to adaptive traffic signal control systems

• AI-based decision-making frameworks for urban mobility

• Reinforcement learning for signal phase and timing optimization

• Integrating predictions into real-time decision engines

• Dashboard interfaces for city traffic managers

• Multi-objective optimization: delay, emissions, throughput

• Case studies: AI-powered signal control in smart cities

• Challenges: scalability, safety, latency, and governance

• AI/ML and computer vision professionals

• Transportation engineers and urban planners

• Researchers in mobility, logistics, and autonomous systems

• Public policy and smart city innovation stakeholders

• Students with technical backgrounds in CS, EE, or civil engineering