About

Zhaodan Kong is an Associate Professor in the Department of Mechanical and Aerospace Engineering at the University of California, Davis. His research interests encompass control theory, machine learning, and formal methods, with a focus on their applications to autonomous systems, human-autonomy teaming, cyber-physical systems, and neural engineering. He leads the Cyber-Human-Physical Systems Lab, where these interdisciplinary areas are explored to advance the understanding and development of complex engineered systems that integrate computational and physical components.

Research topics

• Computer Science
• Artificial Intelligence
• Machine Learning
• Data Mining
• Computer Security
• Engineering
• Mechanical engineering
• Mathematical economics
• Aerospace engineering
• Mathematics
• Control engineering

Selected publications

• Real-Time, Onboard, Model-based Wind Estimation and Control for Multirotor UAVs Flying in Winds

  2026-01-08

  articleSenior author

  This paper presents a real-time, sensor-efficient framework for wind estimation and wind-aware control of multirotor unmanned aerial vehicles (UAVs) operating in strong and persistent winds. The proposed system integrates an enhanced Extended Kalman Filter (ES–EKF), which fuses IMU and GPS data with a high-fidelity 6-DOF aerodynamic model to estimate the horizontal wind vector without requiring dedicated air-data sensors. The estimated wind is then incorporated into a hierarchical control architecture: a nonlinear model predictive controller (NMPC) provides wind-aware, anticipatory feedforward setpoints, while PID controllers supply the high-rate stabilization necessary to reject unmodeled disturbances. High-fidelity simulations demonstrate accurate wind reconstruction and robust trajectory tracking in winds up to 12 m/s, including challenging crosswind scenarios exceeding the vehicle’s nominal cruise velocity. Hardware-in-the-loop (HIL) testing further confirms real-time feasibility of both the estimator and the controller on embedded onboard hardware. Together, these results indicate that the proposed framework enables reliable, wind-aware flight using only standard onboard sensors, and is suitable for deployment in demanding coastal and environmental monitoring missions.

  PublisherDOI

• Utilizing untapped human operator knowledge for smarter and more sustainable manufacturing: examples of unconventional sensor data for the human-in-the-loop integration

  Manufacturing Letters · 2026-05-01

  articleSenior author
  PublisherDOI

• Neural dynamics encoding risky choices during deliberation reveal separate choice subspaces

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-04-30

  preprintOpen access

  Abstract Human decision-making involves the coordinated activity of multiple brain areas, acting in concert, to enable humans to make choices. Most decisions are carried out under conditions of uncertainty, where the desired outcome may not be achieved if the wrong decision is made. In these cases, humans deliberate before making a choice. The neural dynamics underlying deliberation are unknown and intracranial recordings in clinical settings present a unique opportunity to record high temporal resolution electrophysiological data from many (hundreds) brain locations during behavior. Combined with dynamic systems modeling, these allow identification of latent brain states that describe the neural dynamics during decision-making, providing insight into these neural dynamics and computations. Results show that the neural dynamics underlying risky decision, but not decisions without risk, converge to separate subspaces depending on the subject’s preferred choice and that the degree of overlap between these subspaces declines as choice approaches, suggesting a network level representation of evidence accumulation. These results bridge the gap between regression analyses and data driven models of latent states and suggest that during risky decisions, deliberation and evidence accumulation toward a final decision are represented by the same neural dynamics, providing novel insights into the neural computations underlying human choice. Highlights Highly accurate decoding can be accomplished with dynamical systems modeling that revealed distinct attractor-like subspaces, one for each of the two options (gamble or safe bet). Early during deliberation, single trial neural trajectories show rapid transitioning between these subspaces, but as the time of choice selection approached, the single trial trajectories converged towards one of the subspaces, whose identity was consistent with the subject’s choice. These dynamics are specific to risky decisions as we found classification accuracy dropped to chance for trials where the gamble option was guaranteed to be successful (100% win probability) or unsuccessful (0% win probability). These results suggest that deliberation and evidence accumulation toward a final decision in the presence of any risk can be represented by the same neural dynamics, providing novel insights into the neural computations underlying human choice.

  PublisherOA PDFDOI

• An EEG-network-metric based approach to real-time trust inference in human-autonomy teaming

  Frontiers in Neuroergonomics · 2025-09-23 · 3 citations

  articleOpen accessSenior authorCorresponding

  Efficient and effective teaming between humans and autonomous systems requires the establishment and maintenance of trust to maximize team task performance. Despite advances in autonomous systems, human expertise remains critical in tasks fraught with deviations from procedures or plans that cannot be pre-programmed. As autonomous systems become more sophisticated, they will possess the ability to positively influence interactions with their human partners, provided the autonomous systems have a real-time estimation of their human partner's cognitive state (including trust). In this paper, we report our results in ascertaining a human's trust in an autonomous system via electroencephalogram (EEG) measurements. We report that trust can be measured continuously and unobtrusively, and that using analysis techniques which account for interactions among brain regions shows benefits compared to more traditional methods which use only EEG signal-power. Inter-channel connectivity network-metrics, which measure dynamic changes in synchronous behavior between distant brain regions, appear to better capture cognitive activities that correlate with a human's trust in an autonomous system.

  PublisherOA PDFDOI

• Vision-based navigation of unmanned aerial vehicles in orchards: An imitation learning approach

  Computers and Electronics in Agriculture · 2025-08-07 · 5 citations

  articleOpen accessSenior authorCorresponding
  PublisherOA PDFDOI

• Peak Delta Frequency Regulates the Timing of Decisions

  2025-10-26

  article

  Despite extensive study on how the brain decides, the neurophysiological underpinnings of when a decision is made remain poorly understood. Previous modeling work in our lab using human intracranial local field potentials demonstrated that the time at which a decision was made was time-locked to a particular phase of the participants on-going slow wave activity (delta, .5-4Hz). To evaluate whether the on-going delta was driving the time of decision, we repeated our study but forced participants to make a decision at random times throughout their delta cycle. We hypothesized that if delta was driving the time of decision, the phase of the delta prior to decision would be well time locked to the time of decision. Alternatively, if the decision drove the phase of delta by resetting the phase at the time of choice, the time of decision would be time-locked to the phase of the on-going delta after choice. Because we are interested in this low frequency, we could use surface electroencephalogram (EEG). Results showed that participants consistently made their decision during a specific phase of their peak delta (.5-4hz) cycle, regardless of where in their delta cycle they were prompted to make their decision, consistent with our prior study using intracranial local field potentials. Moreover, time of decision was phase-locked to the on-going delta prior to choice selection. Since it has been shown that information about decision making is widely distributed across the brain, these results suggest that delta oscillations could be used to synchronize information from multiple brain regions.

  PublisherDOI

• A Reliable Framework for Human-in-the-Loop Anomaly Detection in Time Series

  ACM Transactions on Interactive Intelligent Systems · 2025-11-27 · 4 citations

  articleSenior author

  Time series anomaly detection is a critical machine learning task for numerous applications, such as finance, healthcare, and industrial systems. However, even high-performing models may exhibit potential issues such as biases, leading to unreliable outcomes and misplaced confidence. While model explanation techniques, particularly visual explanations, offer valuable insights by elucidating model attributions of their decision, many limitations still exist—They are primarily instance-based and not scalable across the dataset, and they provide one-directional information from the model to the human side, lacking a mechanism for users to address detected issues. To fulfill these gaps, we introduce HILAD , a novel framework designed to foster a dynamic and bidirectional collaboration between humans and AI for enhancing anomaly detection models in time series. Through our visual interface, HILAD empowers domain experts to detect, interpret, and correct unexpected model behaviors at scale. Our evaluation through user studies with two models and three time series datasets demonstrates the effectiveness of HILAD , which fosters a deeper model understanding, immediate corrective actions, and model reliability enhancement.

  PublisherDOI

• Implet: A Post-Hoc Subsequence Explainer for Time Series Models

  2025-11-12

  article

  Explainability in time series models is crucial for fostering trust, facilitating debugging, and ensuring interpretability in real-world applications. In this work, we introduce Implet, a novel post-hoc explainer that generates accurate and concise subsequence-level explanations for time series models. Our approach identifies critical temporal segments that significantly contribute to the model's predictions, providing enhanced interpretability beyond traditional feature-attribution methods. Based on it, we propose a cohort-based (group-level) explanation framework designed to further improve the conciseness and interpretability of our explanations. We evaluate Implet on several standard time-series classification benchmarks, demonstrating its effectiveness in improving interpretability. The code is available at https://github.com/LbzSteven/implet.

  PublisherDOI

• CohEx: A Generalized Framework for Cohort Explanation

  Proceedings of the AAAI Conference on Artificial Intelligence · 2025-04-11 · 1 citations

  articleOpen access

  eXplainable Artificial Intelligence (XAI) has garnered significant attention for enhancing transparency and trust in machine learning models. However, the scopes of most existing explanation techniques focus either on offering a holistic view of the explainee model (global explanation) or on individual instances (local explanation), while the middle ground, i.e., cohort-based explanation, is less explored. Cohort explanations offer insights into the explainee's behavior on a specific group or cohort of instances, enabling a deeper understanding of model decisions within a defined context. In this paper, we discuss the unique challenges and opportunities associated with measuring cohort explanations, define their desired properties, and create a generalized framework for generating cohort explanations based on supervised clustering.

  PublisherOA PDFDOI

• Exploring the Suitability of Piecewise-Linear Dynamical System Models for Cognitive Neural Dynamics

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-01-15 · 1 citations

  preprint

  Abstract Dynamical system models have proven useful for decoding the current brain state from neural activity. So far, neuroscience has largely relied on either linear models or nonlinear models based on artificial neural networks. Piecewise linear approximations of nonlinear dynamics have proven useful in other technical applications, providing a clear advantage over network-based models, when the dynamical system is not only supposed to be observed, but also controlled. Here we explore whether piecewise-linear dynamical system models (recurrent Switching Linear Dynamical System or rSLDS models) could be useful for modeling brain dynamics, in particular in the context of cognitive tasks. We first generate artificial neural data based on a nonlinear computational model of perceptual decision-making and demonstrate that piecewise-linear dynamics can be successfully recovered from these observations. We then demonstrate that the piecewise-linear model outperforms a linear model in terms of predicting future states of the system and associated neural activity. Finally, we apply our approach to a publicly available dataset recorded from monkeys performing perceptual decisions. Much to our surprise, the piecewise-linear model did not provide a significant advantage over a linear model for these particular data, although linear models that were estimated from different trial epochs showed qualitatively different dynamics. In summary, we present a dynamical system modeling approach that could prove useful in situations, where the brain state needs to be controlled in a closed-loop fashion, for example, in new neuromodulation applications for treating cognitive deficits. Future work will have to show under what conditions the brain dynamics are sufficiently nonlinear to warrant the use of a piecewise-linear model over a linear one.

  PublisherDOI

Frequent coauthors

• Bernard Mettler

  International Computer Science Institute

  24 shared

• Călin Belta

  Boston University

  10 shared

• Xinfan Lin

  10 shared

• Austin Jones

  9 shared

• John Baillieul

  8 shared

• Nicolas Michel

  8 shared

• Gang Chen

  South China University of Technology

  8 shared

• Chad Goerzen

  7 shared

Education

• PhD, Aerospace Engineering and Mechanics

  University of Minnesota, Twin Cities

  2012

• Master, Astronautics and Mechanics

  Harbin Institute of Technology

  2006

• Bachelor, Astronautics and Mechanics

  Harbin Institute of Technology

  2004

Awards & honors

• CITRIS-CDSS Innovation Fellowship to Advance Early Wildfire…