Junkai Tan谭浚楷
HKPFS Ph.D. Fellow PolyU AAE
Autonomous Control & Safe Learning
I am an HKPFS Ph.D. Fellow in the Department of Aeronautical and Aviation Engineering at The Hong Kong Polytechnic University (PolyU), advised by Prof. Wen-Hua Chen and affiliated with RCLAE.
Current Focus
I am exploring how Prof. Chen’s dual control for exploration and exploitation (DCEE) connects with active inference, studying how robots gather useful information while pursuing a task. My longer-term interests are world models and the links between autonomous control and cognitive neuroscience.
- HKPFS Fellow
- 570 citations (Google Scholar)
- XJTU Student Pacesetter (15 / 30,000+)
Recent Highlights
News archivePolicy–safety separation under polygonal constraints: accepted by SCTS.
Stackelberg attack–defense control: accepted by IEEE/CAA JAS as a regular paper.
Resilient critic-learning control: available in IEEE TII Early Access.
XJTU Graduate with Honor (2025–2026).
Best Student Paper Award, IEEE ICAIS & ISAS · XJTU Qihang Role Model Nominee.
Selected Research
Earlier focus: safe reinforcement learning, prescribed-time control, and differential games, with applications to UAV–UGV coordination and human–UAV interaction.
-
Separating control policy learning from safety enforcement under polygonal constraints
Science China Technological Sciences, 2026.
-
Dual-Clock Guaranteed-Time Actor-Critic Optimal Control with Flexible Input Saturation
IEEE Transactions on Systems, Man, and Cybernetics: Systems, Early Access, pp. 1–14, 2026. IEEE Xplore
-
Fixed-time stochastic learning from human-UAV interaction with state-input constraints
IEEE Transactions on Industrial Electronics, 73(2): 3071–3082, 2026. IEEE Xplore
-
Fixed-time hierarchical game-based unmanned aerial-ground vehicle docking control
IEEE/CAA Journal of Automatica Sinica, 13(5): 1242–1244, 2026. IEEE Xplore
-
Hierarchical safe reinforcement learning control for leader-follower systems with prescribed performance
IEEE Transactions on Automation Science and Engineering, 22: 19568–19581, 2025. IEEE Xplore
Research software
ADP-MATLAB
From data to feedback.
200 critic featuresFixed tanh basis · learned readout
Click a curve to inspect it, or choose a rollout and scrub through time.
Interactive data could not load. The recorded figure and project links remain available.
Inside the experiment
The normalized pendulum uses a data-identified Koopman model and a value critic with 3 quadratic + 197 anchored tanh features. Policy iteration fits the critic’s output weights; the resulting feedback is frozen for these 50 rollouts.
V̂(x) = βᵀφ(x) u(x) = −ĝ(x)ᵀ∇V̂(x) / (2R)
The pulses illustrate signal flow, not measured neuron activations. The plot replays stored states from the 7 September 2026 run, sampled every 0.05 s here (every fifth original sample); no new trajectories or control values are generated.
Education
- 2026 – present
- Ph.D. student, Aeronautical and Aviation EngineeringThe Hong Kong Polytechnic UniversityAdvisor: Prof. Wen-Hua ChenHKPFS & PolyU Presidential Fellowship
- 2023 – 2026
- M.E., Electrical EngineeringXi’an Jiaotong UniversityAdvisors: Prof. Hui Cao & Dr. Shuangsi XueStudent Pacesetter · National & Guodian Nanzi Special Scholarships
- 2019 – 2023
- B.E., Electrical EngineeringXi’an Jiaotong University
Academic Service
Full service record300+ peer reviews recorded on ORCID, including IEEE TAC, IEEE TCNS, IEEE TASE, IEEE TSMC: Systems, and IEEE L-CSS.
Conference reviewer, IEEE CDC · Session Chair, IEEE ICARM 2023.
Contact
Open to collaboration on safe learning, autonomous control, and low-altitude robotics.
junkai.tan@connect.polyu.hk Department of Aeronautical and Aviation EngineeringThe Hong Kong Polytechnic University
Hung Hom, Kowloon, Hong Kong Campus directions