Status: Accepted for presentation at SPIE Defense + Commercial Sensing 2026 (manuscript due April 8, 2026)
Overview
AIRHOUND is an autonomous UAV perception system for detect and pursuit of target drones. I serve as PI on a team of 7 and perception lead, driving the computer vision, sensor fusion, and real-time inference pipeline.
Project Highlights
- RF-DETR object detection optimized with TensorRT for 30ms inference on NVIDIA Jetson Orin, fused with Intel RealSense depth for 3D target localization
- ROS2 autonomy stack: camera driver, detection, depth fusion, and tracking nodes with inter-process communication feeding PX4 offboard control
- Hardware-in-the-loop testing on physical drone platform
- SPIE Defense + Commercial Sensing 2026 — Accepted for presentation; full manuscript due April 2026
- SPARK grant secured for project funding
Technical Stack
| Category | Tools |
|---|---|
| Object Detection | RF-DETR, TensorRT, PyTorch, OpenCV |
| Edge Computing | NVIDIA Jetson Orin NX |
| Depth Sensing | Intel RealSense |
| Robotics | ROS2, PX4 offboard control |
| Languages | Python, C++ |
Team
- Role: Principal Investigator, Perception Lead
- Team Size: 7 members
- Affiliation: Embry-Riddle Aeronautical University
Links
Publication
Malarchick, R. et al. (2026). "Predictive Target Pursuit for Autonomous UAVs using RF-DETR with Depth-Aware State Estimation and Physics-Informed Trajectory Prediction." SPIE Defense + Security 2026 (Poster, April 28, 2026). Manuscript due April 8, 2026.