AI Traffic Analysis and Optimization Course

Lesson 1 • Introduction to Traffic Data Formats

Covers common file formats, schemas, and timestamping conventions used in traffic datasets. Enables students to load and inspect data programmatically.

Lesson 2 • Core Traffic Metrics and KPIs

Defines volume, speed, density, and delay as primary performance indicators. Connects each metric to real-world operational decisions.

Lesson 3 • Data Sources and Collection Methods

Surveys sensor types, probe vehicles, and manual counts as data sources. Prepares students to evaluate data quality before analysis.

Lesson 1 • Feature Engineering for Traffic Data

Creates derived variables such as speed ratios, congestion indices, and lag features. Enriches datasets to improve downstream model performance.

Lesson 2 • Data Fusion and Integration

Merges heterogeneous sources—sensors, GPS, and incident logs—into unified datasets. Resolves spatial and temporal alignment conflicts between sources.

Lesson 3 • Data Aggregation and Resampling

Transforms high-frequency raw counts into analysis-appropriate time intervals. Supports multi-resolution studies from seconds to daily summaries.

Lesson 4 • Handling Missing and Erroneous Data

Identifies causes of missing readings and sensor faults in traffic data. Applies imputation and filtering strategies to restore dataset integrity.

Lesson 1 • Queuing Theory in Traffic Contexts

Applies arrival and service rate models to intersections and bottlenecks. Quantifies delay and queue length under varying demand scenarios.

Lesson 2 • Microscopic and Car-Following Models

Examines individual vehicle behaviour through spacing, reaction time, and acceleration rules. Bridges micro-level dynamics to network-level outcomes.

Lesson 3 • Model Calibration and Validation

Establishes methods for fitting model parameters to observed data and testing predictive accuracy. Ensures models are reliable before AI integration.

Lesson 4 • Macroscopic Flow Theory

Covers the fundamental diagram and conservation equations governing aggregate traffic flow. Provides the theoretical baseline for AI model benchmarking.

Lesson 1 • Time-Series Forecasting Models

Covers ARIMA, exponential smoothing, and hybrid approaches for short-term traffic prediction. Addresses seasonality and non-stationarity common in traffic data.

Lesson 2 • Model Evaluation and Selection

Defines traffic-specific error metrics and model comparison protocols. Guides selection of the best model for deployment based on operational requirements.

Lesson 3 • Supervised Learning Fundamentals

Reviews regression and classification frameworks as applied to traffic prediction tasks. Establishes the train-validate-test workflow used throughout the chapter.

Lesson 4 • Tree-Based and Ensemble Methods

Applies gradient boosting and random forests to speed and volume prediction. Demonstrates feature importance extraction for traffic variable selection.

Lesson 1 • Transformer Models in Traffic Forecasting

Adapts self-attention mechanisms to long-range traffic dependency modelling. Compares transformer performance against recurrent baselines.

Lesson 2 • Recurrent Networks for Temporal Data

Applies LSTM and GRU architectures to sequential traffic time-series prediction. Addresses vanishing gradients and sequence length trade-offs.

Lesson 3 • Graph Neural Networks for Road Networks

Models road networks as graphs and applies GCN and GAT layers to capture spatial dependencies. Enables network-wide simultaneous prediction.

Lesson 4 • Model Interpretability and Explainability

Applies SHAP, attention visualisation, and saliency maps to deep traffic models. Builds practitioner trust and supports regulatory transparency requirements.

Lesson 1 • Object Detection for Vehicles

Trains and deploys YOLO and Faster R-CNN models to detect vehicles in traffic scenes. Evaluates detection accuracy under varying lighting and occlusion conditions.

Lesson 2 • Multi-Object Tracking Algorithms

Implements SORT, DeepSORT, and Kalman filter-based trackers to follow vehicles across frames. Produces trajectory data for speed and flow estimation.

Lesson 3 • Image Processing Fundamentals

Covers preprocessing steps—noise reduction, contrast enhancement, and morphological operations—applied to traffic camera imagery.

Lesson 4 • Incident and Event Detection from Video

Detects stopped vehicles, wrong-way driving, and congestion onset using video analytics. Integrates alerts into traffic management centre workflows.

Lesson 1 • Reinforcement Learning for Signal Control

Formulates intersection control as a Markov decision process and trains RL agents to minimise delay. Covers state space design, reward shaping, and policy evaluation.

Lesson 2 • Signal Control Fundamentals

Reviews fixed-time, actuated, and adaptive signal control strategies and their performance trade-offs. Establishes the baseline against which AI methods are benchmarked.

Lesson 3 • Deployment and Performance Monitoring

Covers field deployment procedures, A/B testing protocols, and continuous performance monitoring for live signal AI systems.

Lesson 4 • Multi-Agent Coordination at Networks

Extends single-intersection RL to coordinated multi-agent systems across arterial corridors. Addresses communication, green-wave progression, and conflict resolution.

Lesson 1 • Incident Management and Response

Integrates AI detection, impact prediction, and automated rerouting to minimise incident-related delay. Covers decision support tools for operators.

Lesson 2 • Strategic Performance Evaluation

Defines network-level KPIs, scenario comparison frameworks, and long-term monitoring protocols for AI traffic systems.

Lesson 3 • Demand Prediction and Management

Forecasts origin-destination demand using ML models and applies demand management strategies to reduce peak congestion.

Lesson 4 • Multimodal Network Coordination

Optimises interactions between road, transit, and active transport modes using shared data platforms. Improves system-wide mobility and reduces single-occupancy vehicle demand.