Research Direction 01
Communication–Control Co-Design
Jointly optimize wireless communication, computation, and control decisions according to control urgency, uncertainty, and risk.
MPCResource AllocationWireless
RESEARCH AGENDA
Research
My research asks how autonomous networked systems can make reliable decisions when communication, computation, energy, and time are all limited.
01 / VISION
Research Vision
I study communication, control, optimization, learning, and security as coupled components of one decision-making system. Rather than optimizing each layer in isolation, my goal is to build methods that understand which information matters to control, which computations are worth performing, and how limited resources should be allocated under uncertainty and attacks.
02 / QUESTIONS
Core Research Questions
Five questions connect most of my current work and define the direction of my doctoral research.
- Q01
How can communication decisions be made control-aware, so that limited wireless resources are allocated according to their actual impact on closed-loop performance rather than conventional communication metrics alone?
- Q02
How can multiple UAVs jointly sense, communicate, plan, and control under bandwidth, energy, sensing, and computing constraints while maintaining safety and scalability?
- Q03
How can learning models such as graph neural networks and large language models accelerate complex optimization while preserving feasibility, safety, and constraint satisfaction?
- Q04
How can communication, control, and resource scheduling be jointly designed so that autonomous networked systems remain safe and operational under cyber attacks, eavesdropping, and model uncertainty?
- Q05
When communication, computation, energy, and time are all scarce, how should an intelligent system determine what information to transmit, what tasks to compute, and which decisions require immediate action?
03 / INTERACTIVE METHOD
From optimization model to deployable AI solver
The animation below shows the workflow I repeatedly use in learning-augmented optimization: formulate the problem, obtain expert solutions with an exact solver, learn the solution map with a neural network, repair the prediction to satisfy hard constraints, and compare it with the optimum.
01
Formulate the optimization problem
Convert a networked-system task into decision variables, objective functions, and hard constraints.
04 / THEMES
Research Themes
The questions above are developed through eight connected research themes, organized into three broader clusters.
Autonomous Networked Systems
Communication, control, mobility, and resilience for connected autonomous systems.
Research Direction 02
Multi-UAV Systems & Autonomous Exploration
Study scalable coordination, mobility, map sharing, and safety-aware planning for teams of autonomous UAVs.
Multi-UAVMappingTrajectory Planning
Research Direction 04
Cyber-Physical Security & Resilience
Model attacks, authentication overhead, and network uncertainty inside control and resource-allocation problems for resilient connected systems.
DDoSRiskResilience
Learning & Optimization
Learning methods that accelerate or structure difficult optimization and control decisions.
Research Direction 03
Learning-Augmented Optimization
Use graph learning, imitation learning, and expert-guided prediction to accelerate difficult combinatorial optimization while preserving feasibility.
GNNMILPImitation Learning
Research Direction 06
LLM-Guided Scheduling & Control
Connect natural-language task descriptions with scheduling, resource allocation, and receding-horizon control for cyber-physical systems.
LLMSchedulingReceding-Horizon Control
Networked Intelligence
Edge intelligence, secure wireless systems, and lightweight sensing under resource constraints.
Research Direction 05
UAV/IoT Edge Computing & VNF Orchestration
Coordinate UAV deployment, service-function placement, task offloading, routing, and energy constraints in edge networks.
Edge ComputingVNFOffloading
Research Direction 07
Secure & Energy-Aware Wireless IoT
Study physical-layer security, energy harvesting, solar-powered networking, and secure data collection in resource-constrained IoT systems.
Physical-Layer SecurityEnergy HarvestingIoT
Research Direction 08
Intelligent Sensing & Lightweight AI
Explore lightweight learning, hierarchical routing, and domain-robust inference for wearable and sensor-driven intelligent systems.
SensorLLMWearable ComputingLightweight AI
05 / METHODS
Methodological Toolbox
I combine model-based optimization and control with learning-based acceleration and networked-system modeling.
Optimization & Control
- Model Predictive Control (MPC)
- Distributed MPC
- MILP / MINLP
- Risk-Aware & Robust Optimization
- Sample Average Approximation
Learning & AI
- Graph Neural Networks
- Imitation Learning
- LLM-Guided Optimization
- DNN Split Inference
- Mixture of Experts
Networked Systems
- Multi-UAV Systems
- Edge Computing & VNF
- Physical-Layer Security
- Energy Harvesting
- 3D Gaussian Mapping