GWU · Washington, DC

The George Washington University

Lead institution. Perception-based drone delivery and landing, AI-based separation assurance, and the security of air traffic control voice communication.

Vision-based landingSeparation assuranceLLMs for air trafficATC voice security

The George Washington University is the lead institution; Principal Investigator Peng Wei leads the eight-institution team. GWU’s research covers aircraft separation assurance for small drones, including heterogeneous fleets, GPS degradation and spoofing, and language-model-based deconfliction; vision-based drone delivery and landing; and the security of air traffic control voice communication. GWU works with Purdue, TSU, UC Irvine, UT Austin, and NVCC, and hosts NVCC students as summer research interns.

Research thrusts

  1. 01

    Perception-based drone delivery

    Benchmarked transformer and CNN detectors for safe autonomous landing with conformal uncertainty, used vision-language models to identify and explain safe landing zones, and is forecasting how long a drop zone will stay clear during tethered delivery.

    UC-01with UC Irvinewith TSU

  2. 02

    Heterogeneous fleets in shared airspace

    Showed that independently trained, attention-based deconfliction policies from different delivery companies can converge to safe co-existence, and analyzed whether the outcome is fair to less capable fleets.

    UC-02with Purdue

  3. 03

    Separation assurance under GPS spoofing

    Robust multi-agent reinforcement learning that models GPS corruption as a zero-sum game, plus runtime safety filtering that protects learned policies during deployment.

    UC-02with UT Austin

  4. 04

    Language models for tactical deconfliction

    A simulation-to-language pipeline in BlueSky and a fine-tuned open language model that makes interpretable deconfliction decisions in closed-loop simulation.

    UC-02

  5. 05

    ATC voice integrity

    A multidimensional evaluation of AI-synthesized controller speech, providing evidence for assessing the risk of unauthorized voice instructions.

    UC-04

Figures

From the research

Pipeline diagram from onboard video through per-agent hazard curves to a calibrated go or wait decision.
GWUForecasting how long a delivery zone will stay clear. Onboard perception tracks people, pets, and objects; a compositional hazard model and a calibrated guarantee decide whether to lower the package or wait.
Two line charts comparing robust and nominal policies as observation contamination increases.
GWU · UT AustinSeparation assurance under increasing GPS corruption. The robust multi-agent policy keeps near mid-air collisions near zero and maintains larger separation than the nominal policy.
Pipeline from BlueSky scenario generation and prompt engineering to LoRA fine-tuning and closed-loop inference.
GWUSimulation-to-language pipeline for fine-tuning a large language model as a tactical deconfliction policy, then deploying it in closed-loop BlueSky simulations.
Training curves showing rewards converging for two companies and collisions falling while successes rise.
GWU · PurdueTwo delivery companies, each training its own attention-based policy, converge to co-existing deconfliction strategies in shared airspace.
Grid of aerial scenes with green and red overlays marking safe and unsafe landing areas at each pipeline stage.
GWUSafe-landing-zone identification for delivery drones, fusing monocular depth with open-vocabulary detection refined by a vision-language model.

Demos

See GWU's work in action

All demos

GWUUC-01Real world

Will the drop zone stay clear?

Overhead drone video of a delivery pad. GWU’s model tracks the people and objects around it, forecasts how long the zone will stay clear, and decides whether to lower the package, wait, or treat the zone as occupied.

GWU · PurdueUC-02Simulation

Two delivery fleets sharing the sky

BlueSky simulation of drones from two delivery companies over Frisco, Texas, each fleet flying its own attention-based multi-agent reinforcement-learning deconfliction policy.

GWUUC-02Simulation

A language model as a deconfliction policy

A fine-tuned large language model issues tactical deconfliction actions to drones in closed-loop BlueSky simulation.

NVCC · GWUUC-05Software

GridEye UTM monitor

A walkthrough of GridEye, the read-only traffic-management dashboard built by NVCC interns. It tracks a 12-drone delivery fleet over simulated Frisco, Texas, and flags flight-boundary, no-fly-zone, and altitude violations.

Use cases

Where GWU contributes

People

Researchers and students

PhD students4

  • Iman Sharifi
  • Mahyar Ghazanfari
  • Aarifah Ullah
  • Dao Vu

Publications

Publications

10 published or accepted papers

  1. A Survey of Security Challenges and Solutions for Advanced Air Mobility and eVTOL Aircraft

    M. Ghazanfari, I. Sharifi, A. Taye, P. Wei, B. C. Ward, X. Koutsoukos, G. Biswas, et al.

    AIAA SciTech Forum, 2026

    ConferenceGWUVanderbilt

  2. A Survey of Security Challenges and Solutions for UAS Traffic Management (UTM) and Small Unmanned Aerial Systems (sUAS)

    I. Sharifi, M. Ghazanfari, A. Taye, P. Wei, M. Ahmed, H. T. Kim, M. Ghasemi, V. Gupta, B. C. Ward, X. Koutsoukos, G. Biswas, et al.

    AIAA SciTech Forum, 2026

    ConferenceGWUPurdueVanderbilt

  3. Fine-Tuning Large Language Models for Cooperative Tactical Deconfliction of Small Unmanned Aerial Systems

    I. Sharifi, A. Zongo, P. Wei

    CVPR Workshop on Multi-Agent Embodied Intelligent Systems Meet Generative-AI Era (MEIS), 2026

    WorkshopGWUUC-02

  4. Generating Realistic Air Traffic Control Voice Communication using AI-based Text-to-Speech Models

    D. Vu, P. Wei

    Integrated Communications, Navigation and Surveillance Conference (ICNS), 2026

    ConferenceGWUUC-04

  5. Robust Multi-Agent Reinforcement Learning for Small UAS Separation Assurance under GPS Degradation and Spoofing

    A. Zongo, F. Fotiadis, U. Topcu, P. Wei

    IEEE Conference on Decision and Control (CDC), 2026Accepted

    ConferenceGWUUT AustinUC-02

  6. Runtime Safety Filtering for Learned Small UAS Separation Policies under GNSS Degradation

    A. Zongo, P. Wei

    AIAA/IEEE Digital Avionics Systems Conference (DASC), 2026

    ConferenceGWUUC-02

  7. See&Say: Vision Language Guided Safe Zone Detection for Autonomous Package Delivery Drones

    M. Ghazanfari, P. Wei

    AIAA/IEEE Digital Avionics Systems Conference (DASC), Orlando, FL, 2026

    ConferenceGWUUC-01

  8. Separation Assurance between Heterogeneous Fleets of Small Unmanned Aerial Systems via Multi-Agent Reinforcement Learning

    I. Sharifi, H. T. Kim, M. H. Ahmed, M. Ghasemi, P. Wei

    IEEE International Conference on Intelligent Transportation Systems (ITSC), 2026

    ConferenceGWUPurdueUC-02

  9. Towards Automated Air Traffic Safety Assessment Around Non-Towered Airports Using Large Language Models

    T. Darrell, M. Ghazanfari, J. Kam, A. Bayen, A. Tabrizian, P. Wei

    AIAA Aviation Forum, 2026

    ConferenceGWU

  10. Transformer or CNN? Benchmarking Real-Time Detection Transformer and YOLOv8 for Small UAS Autonomous Landing

    C. Vu, M. Ghazanfari, K. Dong, A. Taye, A. Tabrizian, P. Wei

    AIAA Aviation Forum, 2025

    ConferenceGWUUC-01