본문 바로가기
  • Home

드론의 자율 주행 제어를 위한 AI 기반의 속도 제어 학습 모델 설계 및 학습 파이프 라인 구축

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(6), pp.101~108
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : May 8, 2025
  • Accepted : June 17, 2025
  • Published : June 30, 2025

Woon-Yong Kim 1

1강원도립대학교

Accredited

ABSTRACT

In this paper, we propose an AI-based learning pipeline for velocity-based autonomous drone navigation. The proposed method utilizes a supervised learning approach that integrates the drone's current state information (velocity and orientation) with future path data to predict velocity commands in real-time. The hybrid LSTM-based model classifies diverse flight patterns such as hovering, straight flight, turning, ascending/descending, obstacle avoidance, and takeoff/landing, generating adaptive velocity commands suitable for different flight scenarios. Consequently, this method enables smooth and efficient path following in complex and dynamic environments. Evaluated within ROS 2 and PX4 simulation environments, the proposed model demonstrates excellent performance metrics, including RMSE of 0.0347, MAE of 0.0225, and an R² of 98.67%. By recognizing various drone flight modes and dynamically calculating appropriate velocities based on current position and speed, the method ensures natural trajectory control and responsive maneuverability.

Citation status

* References for papers published after 2023 are currently being built.