# Chengyuan Zhang > Postdoctoral Researcher in Civil Engineering at McGill University, working with > Prof. Lijun Sun. Research on stochastic models of human driving behavior, > Bayesian inference and calibration, and uncertainty-aware traffic simulation, > from individual decisions to network-scale dynamics. ORCID > 0000-0001-8463-7380. Contact: > enzozcy@gmail.com. The research guides below are the best entry points: each one states a question and answers it directly, then links to the papers, code, and notes behind the answer. Dates and venues on the publications page are authoritative; the research summaries are written for readers, not as a substitute for the papers. ## Research guides - [Bayesian Calibration and Inference for Car-Following Models](https://chengyuan-zhang.github.io/research/bayesian-calibration/): How Bayesian car-following calibration connects trajectory data, hierarchical parameters, temporal residuals, posterior checks, and public IDM implementations. - [From Driver Variability to Traffic Simulation](https://chengyuan-zhang.github.io/research/traffic-simulation/): Explore how car-following dynamics and temporally structured driver variability enter traffic simulation, with research papers, code, and an interactive ring-road model. - [Stochastic Driving Behavior: Memory, Heterogeneity, and Regimes](https://chengyuan-zhang.github.io/research/stochastic-driving-behavior/): A research guide to temporal memory, differences between drivers, latent behavioral regimes, and unexplained variability in car-following models. ## Paper summaries - [Bayesian Calibration of the Intelligent Driver Model](https://chengyuan-zhang.github.io/papers/bayesian-idm-calibration/): Hierarchical Bayesian inference represents uncertainty in IDM parameters, while a Gaussian process models temporally correlated discrepancies in memory-augmented IDM. - [Calibrating Car-Following Models via Bayesian Dynamic Regression](https://chengyuan-zhang.github.io/papers/bayesian-dynamic-regression/): Dynamic regression combines a parsimonious car-following model with autoregressive error dynamics, retaining historical information for calibration and probabilistic simulation. - [Modeling Within-Driver Heterogeneity in Car-Following through Coupled Response–Scenario Dynamics](https://chengyuan-zhang.github.io/papers/markov-regime-switching-idm/): PS-HMM-IDM models driving response regimes and traffic scenarios using a joint Markov process and regime-specific IDM response relationships. - [When Context Is Not Enough: Modeling Unexplained Variability in Car-Following Behavior](https://chengyuan-zhang.github.io/papers/when-context-is-not-enough/): Context-dependent predictions and a nonstationary Gaussian process jointly represent the expected acceleration and the evolving temporal structure of residual variability. ## Main pages - [Home](https://chengyuan-zhang.github.io/): Dr. Chengyuan Zhang — Postdoctoral Researcher in Civil Engineering, McGill University (with Prof. Lijun Sun). Stochastic models of human driving behavior, Bayesian inference and calibration, and uncertainty-aware traffic simulation from individual decisions to network-scale dynamics. - [Publications](https://chengyuan-zhang.github.io/publications/): complete list of journal articles, conference papers, books, and patents, with DOIs, research summaries, code, and BibTeX. - [Code & Data](https://chengyuan-zhang.github.io/code/): public implementations behind the papers, plus an in-browser ring-road traffic simulator. - [Curriculum Vitae](https://chengyuan-zhang.github.io/cv/): education, positions, fellowships, awards, and academic service. - [Talks](https://chengyuan-zhang.github.io/talks/): invited lectures and conference presentations. - [News](https://chengyuan-zhang.github.io/news/): recent talks, publications, and research updates. - [Teaching](https://chengyuan-zhang.github.io/teaching/): teaching assistant experience at McGill University. ## Research notes - [Hidden Markov Model and Driving Behavior Modeling: From HMMs to Factorial HMMs to FHMM–IDM --- a three–part primer](https://chengyuan-zhang.github.io/posts/FHMM-IDM/) (2025-08-14): A three-part primer running from hidden Markov models to factorial HMMs to FHMM-IDM, an interpretable regime-switching model of car-following behavior. - [Introduction to Autoregressive (AR) Processes](https://chengyuan-zhang.github.io/posts/AR/) (2025-02-27): An introduction to autoregressive processes: mathematical formulation, stationarity, and the AR model variants used in time-series forecasting. - [Fundamental Probabilistic Graphical Models: Tail-to-Tail, Head-to-Tail, and Head-to-Head](https://chengyuan-zhang.github.io/posts/three-PGMs/) (2025-01-25): The three elementary structures in directed graphical models, and how each one decides when two variables are conditionally independent. - [A Detailed Introduction to Gaussian Velocity Fields (GVF) Based on Gaussian Processes](https://chengyuan-zhang.github.io/posts/GVF/) (2025-01-10): Gaussian velocity fields represent surrounding traffic at fixed dimensionality, so the model does not depend on how many vehicles happen to be nearby. - [Gaussian Processes (GP) for Time Series Forecasting](https://chengyuan-zhang.github.io/posts/gp-time-series/) (2024-12-22): Using Gaussian processes for time-series forecasting: the modeling framework, kernel choices, and the predictive uncertainty that point forecasts leave out. - [Connections among autoregressive (AR) processes, Cochrane-Orcutt correction, Ornstein-Uhlenbeck (OU) processes, and Gaussian Processes (GP)](https://chengyuan-zhang.github.io/posts/processes_connections/) (2024-12-20): How autoregressive processes, the Cochrane-Orcutt correction, Ornstein-Uhlenbeck processes, and Gaussian processes relate, with car-following applications. - [Modeling Autocorrelation: FFT vs Gaussian Processes](https://chengyuan-zhang.github.io/posts/autocorrelation/) (2024-12-18): Fourier transforms and Gaussian processes model autocorrelation in the frequency and time domains; the spectral representation theorem connects the two. - [From Ordinary Least Squares (OLS) to Generalized Least Squares (GLS)](https://chengyuan-zhang.github.io/posts/ols-to-gls/) (2024-12-17): Why OLS remains unbiased but loses efficiency under heteroskedasticity or correlated errors, and how generalized least squares restores it. - [Proof: unbiasedness of ordinary least squares (OLS)](https://chengyuan-zhang.github.io/posts/ols-unbiased/) (2024-12-17): A step-by-step proof that the ordinary least squares estimator is unbiased under the standard linear regression assumptions. - [Random Effects and Hierarchical Models in Driving Behaviors Modeling](https://chengyuan-zhang.github.io/posts/random-effects/) (2024-11-27): Random effects and hierarchical models for driver heterogeneity: representing how IDM parameters differ across drivers instead of assuming one shared set. - [Heterogeneity and Hierarchical Models: Understanding Pooled, Unpooled, and Hierarchical Approaches](https://chengyuan-zhang.github.io/posts/hierarchical/) (2024-11-24): How hyperparameters distinguish pooled, unpooled, and hierarchical (partially pooled) models, and what each choice implies for flexibility and interpretability. - [Understanding the Log-Sum-Exp Trick and Its Application in Hidden Markov Models (HMMs)](https://chengyuan-zhang.github.io/posts/logsumexp/) (2024-11-20): What the log-sum-exp trick is, why numerical underflow makes it necessary, and how it is applied in the forward-backward algorithm for hidden Markov models. - [Matrix derivative of Frobenius norm involving Hadamard product](https://chengyuan-zhang.github.io/posts/matrix-derivative/) (2022-10-01): A worked derivation of the matrix derivative of a Frobenius norm involving a Hadamard product, taken with respect to both factors of the inner product. ## Optional - [Sitemap (XML)](https://chengyuan-zhang.github.io/sitemap.xml): every indexable URL on the site. - [Atom feed](https://chengyuan-zhang.github.io/feed.xml): full text of the research notes. - [Ring-road traffic simulator](https://chengyuan-zhang.github.io/traffic-sim/): interactive demonstration of car-following dynamics and stop-and-go waves. - [Miscellaneous](https://chengyuan-zhang.github.io/miscellaneous/): photography and other non-research pages.