Research on UAV Autonomous Trajectory Planning Based on Prediction Information in Crowded Unknown Dynamic...

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Vision

Background:

• Unmanned aerial vehicles (UAVs) face challenges in autonomous ultra-low-altitude flight due to unpredictable crowd dynamics.
• Dense crowds present complex, dynamic obstacles that threaten UAV safety and operational feasibility.

Purpose of the Study:

• To develop an integrated UAV trajectory planning method for safe autonomous flight in crowded environments.
• To enhance UAV navigation by combining accurate pedestrian trajectory prediction with robust gradient-based planning.

Main Methods:

• A Contrastive Distribution Latent Code Generator (CDLCG) was developed for pedestrian trajectory prediction, inferring future distributions from historical data.
• An adaptive gradient-based UAV trajectory planning approach was designed, incorporating adaptive cost weights for different optimization stages and obstacle types.
• Validation was performed using public datasets, OptiTrack Motion Capture System experiments, and Gazebo simulations with varying crowd densities.

Main Results:

• The CDLCG model accurately predicted pedestrian trajectories, validated through simulations and physical experiments.
• The adaptive gradient-based planning method significantly improved the success rate of UAV trajectory planning in dynamic, crowded environments.
• The proposed method effectively balanced trajectory smoothness, safety, and feasibility, ensuring secure UAV operations.

Conclusions:

• The integrated approach provides a reliable solution for autonomous UAV navigation in complex, dynamic crowd scenarios.
• This research contributes to safer and more efficient UAV operations in challenging real-world environments.
• The developed methods enhance the capability of UAVs to operate autonomously and safely amidst human crowds.