Sensor-Model-Based Trajectory Optimization for UAVs to Enhance Detection Performance: An Optimal Control Approach and Experimental Results | JoVE Visualize

Area of Science:

Robotics and Automation
Artificial Intelligence
Remote Sensing

Background:

Unmanned aerial vehicles (UAVs) are crucial for aerial reconnaissance, requiring high object detection accuracy for mission success.
Environmental conditions significantly impact sensor data acquisition and automated object detection performance.
Existing methods lack a comprehensive approach to mitigate environmental effects on UAV-based detection.

Purpose of the Study:

To develop and evaluate a novel sensor performance model for mapping environmental influences on detection accuracy.
To implement sensor-model-based trajectory optimization for fixed-wing UAVs to enhance detection performance.
To integrate deep learning-based object detection within a perception chain for UAV reconnaissance.

Main Methods:

A new sensor performance model was developed to quantify the impact of environmental states on detection performance.
Nonlinear model predictive control (NMPC) and dynamic programming were employed for trajectory optimization.
Optimized reference flight trajectories were calculated, aligning UAV and sensor positioning with reconnaissance targets.
Constraints including perceptual, platform-specific, environmental, and mission requirements were incorporated into the optimization.

Main Results:

The developed sensor performance model is the first to map detection performance for a deep learning object detector concerning environmental states in UAV reconnaissance.
Sensor-model-based trajectory optimization using NMPC achieved an average 4.48% increase in detection performance compared to benchmarks.
Dynamic programming yielded detection performance values equal to or closely approaching theoretical maximums.

Conclusions:

The proposed sensor performance model and trajectory optimization approach effectively enhance UAV-based object detection accuracy.
Optimizing flight trajectories based on environmental conditions is a viable strategy to improve aerial reconnaissance mission success.
This work provides a foundational framework for intelligent sensor-model-based trajectory planning in UAV applications.