I am currently a Research Fellow at Monash University, working on
UAV algorithm design and intelligent autonomous systems.
I am also an Honorary Research Fellow at the University of Melbourne, where I supervise
student research projects. I received my PhD in Artificial Intelligence
from the University of Melbourne, with a thesis titled
Planning and Goal Recognition in Humans and Machines.
My research focuses on developing computational models that help AI
systems better understand human mental states, such as
intentions, beliefs, and goals,
to enable more effective interaction with people. By drawing on methods from
automated planning, robotics, human-AI interaction, cognitive science, and statistics,
I aim to build AI systems that are better aligned with human values and needs,
while also being more interpretable and explainable. My work has been
published in leading venues including
ICAPS, AAMAS, KR, CHI, CogSci, and
IEEE Robotics and Automation Letters, and I actively contribute to
the research community through conference organisation and reviewing, including service
for ICAPS, CogSci, IJCAI, ICRA, KR, and Robotics and Autonomous Systems.
Honoured to be selected as a DAAD AINeT Fellow on AI in Neuroscience. The fellowship supports international networking and research collaboration between postdoctoral researchers in AI and neuroscience, with opportunities for visiting research stays at German institutions.
Apr. 2026
Our paper A Probabilistic Framework for Hierarchical Goal Recognition has been accepted at KR 2026 — the 23rd International Conference on Principles of Knowledge Representation and Reasoning. We propose a probabilistic framework that extends goal recognition to hierarchical goal structures.
Apr. 2026
Excited to be co-organising two workshops at ICAPS 2026 (Dublin, Ireland, June 27–July 2)! We are running HAXP (Human-Aware and Explainable Planning) and PR-BGI (Planning and Reasoning about Beliefs, Goals and Intentions). Both workshops bring together cutting-edge research on human-AI interaction and autonomous agent reasoning — submissions are welcome!
Feb. 2026
Our paper Proactive Assistance Agent with Online Goal Recognition has been accepted at ICAPS 2026 — the 36th International Conference on Automated Planning and Scheduling. This work addresses proactive assistance by combining online goal recognition with planning in goal-directed environments.
Jan. 2026
Our paper Neurosymbolic Active Goal Recognition in Partially Observable Environments has been accepted at AAMAS 2026 — the 25th International Conference on Autonomous Agents and Multiagent Systems. This work extends active goal recognition into partially observable settings using a neuro-symbolic approach.
Jul. 2025
Our paper Probabilistic Active Goal Recognition has been accepted at KR 2025 — the 22nd International Conference on Principles of Knowledge Representation and Reasoning. We present a probabilistic framework combining joint belief updates with Monte Carlo Tree Search for efficient active goal inference in multi-agent settings.
Jun. 2025
Our paper NEUSIS has been accepted in IEEE Robotics and Automation Letters. NEUSIS is a compositional neuro-symbolic framework that integrates visual perception, probabilistic world modelling, and hierarchical planning for autonomous UAV search missions in complex real-world environments.
Jun. 2025
Started a new position as Research Fellow (Level B) at Monash University, joining the HARNESS project (Hierarchical Abstractions and Reasoning for Neuro-Symbolic Systems). I am working on UAV algorithm design, developing hierarchical symbolic algorithms for autonomous navigation and search in unknown environments.
Apr. 2025
Our paper on modeling human sequential decision-making in the Tower of London has been accepted at CogSci 2025 — the 47th Annual Meeting of the Cognitive Science Society. We introduce a profile-based framework and a replanning detection component leveraging thinking time to model individual differences in planning behaviour.
Feb. 2025
Our paper Modeling Higher-Order Human Beliefs Using the Justified Perspective Model has been accepted at CHI 2025 Extended Abstracts — the ACM CHI Conference on Human Factors in Computing Systems, Yokohama, Japan. This work explores the feasibility of anticipating higher-order human beliefs using the Justified Perspective model.
Jun. 2024
Started as Research Fellow (Level B) at Monash University on the ARC discovery project Human Models for Accelerated Robot Learning and Human-Robot Interaction, and concurrently as Honorary Research Fellow at the University of Melbourne, where I supervise student research projects.
May 2024
Honoured to receive the Pragnesh Jay Modi Best Paper Award at AAMAS 2024 (Auckland, NZ) for our paper Human Goal Recognition as Bayesian Inference. This award recognises the best paper at the premier conference on autonomous agents and multi-agent systems.
May 2024
Successfully defended my PhD thesis, Planning and Goal Recognition in Humans and Machines, at the University of Melbourne. Grateful to my supervisors Associate Professor Nir Lipovetzky and Professor Charles Kemp for their guidance throughout my doctoral journey.
Dec. 2023
Our paper Human Goal Recognition as Bayesian Inference has been accepted at AAMAS 2024 — the 23rd International Conference on Autonomous Agents and Multiagent Systems. We use a Bayesian framework to show that actions, timing, and goal solvability all play a role in how humans recognise others' goals.
Apr. 2023
Our paper Comparing AI Planning Algorithms With Humans on the Tower of London Task has been accepted at CogSci 2023 — the 45th Annual Meeting of the Cognitive Science Society. We show that an adaptive lookahead planner provides the best account of human actions and planning times on this classic task.
Feb. 2023
Our paper Goal Recognition With Timing Information has been accepted at ICAPS 2023 — the 33rd International Conference on Automated Planning and Scheduling. We show that incorporating action timing substantially improves goal recognition, especially when few actions have been observed.
Publications
2026
Neurosymbolic Active Goal Recognition in Partially Observable Environments
25th International Conference on Autonomous Agents and Multiagent Systems AAMAS 2026CORE A*
Chenyuan Zhang, Sukai Huang, Hamid Rezatofighi, Mor Vered, Buser Say 1st author
Proactive Assistance Agent with Online Goal Recognition
36th International Conference on Automated Planning and Scheduling ICAPS 2026CORE A*
In multi-agent environments, effective interaction hinges on understanding the beliefs and intentions of other agents. While prior work on goal recognition has largely treated the observer as a passive reasoner, Active Goal Recognition (AGR) focuses on strategically gathering information to reduce uncertainty. We adopt a probabilistic framework for AGR and propose an integrated solution that combines a joint belief update mechanism with a Monte Carlo Tree Search (MCTS) algorithm, allowing the observer to plan efficiently and infer the actor's hidden goal without requiring domain-specific knowledge. Through comprehensive empirical evaluation in a grid-based domain, we show that our joint belief update significantly outperforms passive goal recognition, and that our domain-independent MCTS performs comparably to our strong domain-specific greedy baseline. These results establish our solution as a practical and robust framework for goal inference, advancing the field toward more interactive and adaptive multi-agent systems.
@inproceedings{zhang2025probabilistic,
title = {Probabilistic Active Goal Recognition},
author = {Zhang, Chenyuan and Rojas Cardenas, Cristian and Rezatofighi, Hamid and Vered, Mor and Say, Buser},
booktitle = {Proceedings of the 22nd International Conference on Principles of Knowledge Representation and Reasoning},
pages = {880--890},
year = {2025},
doi = {10.24963/kr.2025/85}
}
Modeling Human Sequential Decision-Making in the Tower of London: Incorporating Individual Differences and Timing-Based Replanning Inference
Chenyuan Zhang, Yuansan Liu, Dana Kulić, Pamela Carreno-Medrano, Michael Burke
47th Annual Meeting of the Cognitive Science Society CogSci 2025CORE B
Chenyuan Zhang, Yuansan Liu, Dana Kulić, Pamela Carreno-Medrano, Michael Burke 1st author
Modeling human sequential decision-making behavior presents a significant challenge for researchers in artificial intelligence, robotics, and cognitive science. In this paper, we introduce a human behavior model designed to predict actions in the Tower of London task, addressing two critical aspects that have been largely overlooked in existing methodologies. First, we propose a profile-based action prediction framework that extracts user and task profiles from historical data, enhancing action prediction in novel scenarios. Second, we introduce a replanning detection component that leverages thinking time as an indicator of planning processes in the human mind, enabling a more precise representation of cognitive dynamics. Our evaluations demonstrate the effectiveness of the proposed model, achieving superior performance in behavior prediction within the Tower of London task. This work lays the foundation for more robust human behavior modeling in sequential decision-making environments.
@inproceedings{zhang2025modeling,
title = {Modeling Human Sequential Decision-Making in the Tower of London: Incorporating Individual Differences and Timing-Based Replanning Inference},
author = {Zhang, Chenyuan and Liu, Yuansan and Kuli\'{c}, Dana and Carreno-Medrano, Pamela and Burke, Michael},
booktitle = {Proceedings of the 47th Annual Meeting of the Cognitive Science Society},
year = {2025}
}
Modeling Higher-Order Human Beliefs Using the Justified Perspective Model
This work presents a pilot study involving five participants to explore the feasibility of anticipating higher-order human beliefs through a generalizable formalization using the Justified Perspective (JP) model, and introduces a novel processing algorithm inspired by Item Response Theory (IRT) to effectively estimate human belief reasoning ability from their responses across different scenarios. The preliminary results suggest that human reasoning ability is consistent within the JP model framework, and is positively correlated with nesting ability. The findings highlight the potential of bridging theoretical epistemic frameworks with practical human-centered applications, providing valuable insights for the design of intelligent human-computer interaction systems that more effectively comprehend human cognition.
@inproceedings{li2025modeling,
title = {Modeling Higher-Order Human Beliefs Using the Justified Perspective Model},
author = {Li, Wanchun and Zhang, Chenyuan and Li, Weijia and Hu, Guang and Xu, Yangmengfei},
booktitle = {Extended Abstracts of the CHI Conference on Human Factors in Computing Systems},
year = {2025},
doi = {10.1145/3706599.3720223}
}
NEUSIS: A Compositional Neuro-Symbolic Framework for Autonomous Perception, Reasoning, and Planning in Complex UAV Search Missions
Zhixi Cai, Cristian Rojas Cardenas, Kevin Leo, Chenyuan Zhang, et al.
IEEE Robotics and Automation Letters RA-L 2025Q1 Journal
Zhixi Cai, Cristian Rojas Cardenas, Kevin Leo, Chenyuan Zhang, et al. co-1st author
This paper addresses the problem of autonomous UAV search missions, where a UAV must locate specific Entities of Interest (EOIs) within a time limit, based on brief descriptions in large, hazard-prone environments with keep-out zones. The UAV must perceive, reason, and make decisions with limited and uncertain information. We propose NEUSIS, a compositional neuro-symbolic system designed for interpretable UAV search and navigation in realistic scenarios. NEUSIS integrates neuro-symbolic visual perception, reasoning, and grounding (GRiD) to process raw sensory inputs, maintains a probabilistic world model for environment representation, and uses a hierarchical planning component (SNaC) for efficient path planning. Experimental results from simulated urban search missions using AirSim and Unreal Engine show that NEUSIS outperforms a state-of-the-art vision-language model and a state-of-the-art search planning model in success rate, search efficiency, and 3D localization. These results demonstrate the effectiveness of our compositional neuro-symbolic approach in handling complex, real-world scenarios, making it a promising solution for autonomous UAV systems in search missions.
@article{cai2025neusis,
title = {NEUSIS: A Compositional Neuro-Symbolic Framework for Autonomous Perception, Reasoning, and Planning in Complex UAV Search Missions},
author = {Cai, Zhixi and Rojas Cardenas, Cristian and Leo, Kevin and Zhang, Chenyuan and Backman, Kal and Li, Hanbing and Li, Boying and Ghorbanali, Mahsa and Datta, Stavya and Qu, Lizhen and Gutierrez Santiago, Julian and Ignatiev, Alexey and Li, Yuan-Fang and Vered, Mor and Stuckey, Peter J. and Garcia de la Banda, Maria and Rezatofighi, Hamid},
journal = {IEEE Robotics and Automation Letters},
volume = {10},
number = {9},
pages = {9502--9509},
year = {2025}
}
2024
Human Goal Recognition as Bayesian Inference: Investigating the Impact of Actions, Timing, and Goal Solvability
23rd International Conference on Autonomous Agents and Multiagent Systems AAMAS 2024CORE A*
Chenyuan Zhang, Charles Kemp, Nir Lipovetzky 1st author
Goal recognition is a fundamental cognitive process that enables individuals to infer intentions based on available cues. Current goal recognition algorithms often take only observed actions as input, but here we use a Bayesian framework to explore the role of actions, timing, and goal solvability in goal recognition. We analyze human responses to goal-recognition problems in the Sokoban domain, and find that actions are assigned most importance, but that timing and solvability also influence goal recognition in some cases, especially when actions are uninformative. We leverage these findings to develop a goal recognition model that matches human inferences more closely than do existing algorithms. Our work provides new insight into human goal recognition and takes a step towards more human-like AI models.
@inproceedings{zhang2024human,
title = {Human Goal Recognition as Bayesian Inference: Investigating the Impact of Actions, Timing, and Goal Solvability},
author = {Zhang, Chenyuan and Kemp, Charles and Lipovetzky, Nir},
booktitle = {Proceedings of the 23rd International Conference on Autonomous Agents and Multiagent Systems},
year = {2024},
doi = {10.5555/3635637.3663071}
}
2023
Goal Recognition With Timing Information
Chenyuan Zhang, Charles Kemp, Nir Lipovetzky
33rd International Conference on Automated Planning and Scheduling ICAPS 2023CORE A*
Chenyuan Zhang, Charles Kemp, Nir Lipovetzky 1st author
Goal recognition has been extensively studied by AI researchers, but most algorithms take only observed actions as input. Here we argue that the time taken to carry out these actions provides an additional signal that supports goal recognition. We present a behavioral experiment confirming that people use timing information in this way, and develop and evaluate a goal recognition algorithm that is sensitive to both actions and timing information. Our results suggest that existing goal recognition algorithms can be improved by incorporating a model of planning time on both synthetic data and human data, and that these improvements can be substantial in scenarios in which relatively few actions have been observed.
@article{zhang2023goal,
title = {Goal Recognition with Timing Information},
author = {Zhang, Chenyuan and Kemp, Charles and Lipovetzky, Nir},
journal = {Proceedings of the International Conference on Automated Planning and Scheduling},
volume = {33},
number = {1},
pages = {443--451},
year = {2023},
doi = {10.1609/icaps.v33i1.27224}
}
Comparing AI Planning Algorithms With Humans on the Tower of London Task
Chenyuan Zhang, Nir Lipovetzky, Charles Kemp
45th Annual Meeting of the Cognitive Science Society CogSci 2023CORE A
Chenyuan Zhang, Nir Lipovetzky, Charles Kemp 1st author
This paper explores the extent to which modern AI planners can account for human performance on the Tower of London (TOL) task, a sequential planning puzzle extensively studied in cognitive psychology. We characterize the task using the Planning Domain Definition Language (PDDL) and evaluate an adaptive online planner alongside a family of optimal and satisficing planners. Each planner is assessed on its ability to predict the actions and planning times of participants in a new behavioral experiment. The results suggest that participants use a range of strategies but that an adaptive lookahead planner provides the best overall account of both human actions and planning times.
@inproceedings{zhang2023comparing,
title = {Comparing AI Planning Algorithms With Humans on the Tower of London Task},
author = {Zhang, Chenyuan and Lipovetzky, Nir and Kemp, Charles},
booktitle = {Proceedings of the 45th Annual Meeting of the Cognitive Science Society},
year = {2023}
}
Teaching
Monash University 2025–present
FIT5047 Fundamentals of Artificial Intelligence
University of Melbourne 2020–2024
COMP90054 AI Planning for Autonomy
COMP30027 Machine Learning
COMP90038 Algorithms and Complexity
COMP30024 Artificial Intelligence
COMP30026 Models of Computation
Awards & Honors
2024
Pragnesh Jay Modi Best Paper Award, AAMAS
2023
Melbourne School of Engineering Travelling Scholarship
2023
Engineering and IT Conference Travel Scholarship
2020–2023
Melbourne Research Scholarship
2018, 2019
Dean's Honours List, University of Melbourne
