AttackIntentInference

Attack Intent Inference of Hypersonic Glide Vehicle based on a Unified Dynamics and Decision-making Model

Overall Architecture

This repository implements a unified framework for inferring the most probable attack target of a Hypersonic Glide Vehicle (HGV). The framework combines dynamics-based trajectory propagation and decision-making models under uncertainty.

📝 Citation

Y. Nam, H. Lee, H. Choi, W.-S. Ra, and C. Kwon,
"Attack Intent Inference of Hypersonic Glide Vehicle based on a Unified Dynamics and Decision-Making Model", IEEE Transactions on Aerospace and Electronic Systems, 2025.

📂 Project Structure

File	Description
main.py	Runs simulation episodes
inference.py	Infers the HGV's target intent using a Bayesian framework
ckf.py	Implements the Cubature Kalman Filter
model.py	Defines the physical dynamics model of the Hypersonic Glide Vehicle (HGV)
model_jax.py	Provides JAX-based rollout functions for efficient trajectory propagation
plot.py	Contains 3D visualization utilities for trajectories
utils.py	Includes helper functions, constants, and configuration parameters

⚙️ Environment Requirements

Tested on:

• Python: 3.10.16
• CUDA 12, cuDNN 9.8
• GPU Acceleration: JAX with jax-cuda12-pjrt

To replicate the environment:

conda env create -f intent.yaml
conda activate intent

🚀 How to Run

python main.py

This command runs user-defined independent simulations, each with randomized threat levels and initial states for the Hypersonic Glide Vehicle (HGV).

The number of iterations is controlled by the loop at the end of main.py (e.g., for itr in range(20)).

The results are saved only when the variable "save_result = True".

Each output is stored as .npz files in the specified directory ./[FolderName]/, where FolderName is passed as an argument in test (foldername='Result', iteration=itr).