UAV Active Perception and Motion Control for Improving Navigation Using Low-Cost Sensors (2407.15122v1)

Abstract: In this study a model pipeline is proposed that combines computer vision with control-theoretic methods and utilizes low cost sensors. The proposed work enables perception-aware motion control for a quadrotor UAV to detect and navigate to objects of interest such as wind turbines and electric towers. The distance to the object of interest was estimated utilizing RGB as the primary sensory input. For the needs of the study, the Microsoft AirSim simulator was used. As a first step, a YOLOv8 model was integrated providing the basic position setpoints towards the detection. From the YOLOv8 inference, a target yaw angle was derived. The subsequent algorithms, combining performant in computational terms computer vision methods and YOLOv8, actively drove the drone to measure the height of the detection. Based on the height, an estimate of the depth was retrieved. In addition to this step, a convolutional neural network was developed, namely ActvePerceptionNet aiming at active YOLOv8 inference. The latter was validated for wind turbines where the rotational motion of the propeller was found to affect object confidence in a near periodical fashion. The results of the simulation experiments conducted in this study showed efficient object height and distance estimation and effective localization.

• The paper presents a model pipeline enhancing UAV active perception and motion control using low-cost RGB sensors and advanced computer vision.
• The system architecture integrates YOLOv8, a novel ActivePerceptionNet CNN, and an Extended Kalman Filter for robust object detection, localization, and navigation.
• Simulation results demonstrate accurate height/depth estimation (e.g., 0.06m error for wind turbines) and increased detection confidence, showing potential for cost-efficient real-world applications.

UAV Active Perception and Motion Control Using Low-Cost Sensors

This paper presents a sophisticated model pipeline that enhances the active perception and motion control of Unmanned Aerial Vehicles (UAVs) specifically integrating low-cost sensory systems. The paper is centered on deploying a computer vision-based framework utilizing low-cost RGB sensors integrated through the Microsoft AirSim simulator to improve the navigation of UAVs towards distinct objects, such as wind turbines and electric towers. The authors propose an intricate combination of the YOLOv8 object detection model and a novel CNN architecture, ActivePerceptionNet, to advance perception-aware UAV motion control.

Technical Implementation

The developed system architecture divides into three main modules: Object Tracking, Extended Kalman Filter (EKF) Localization, and Planning and Control. The primary sensory data stems from RGB imagery, IMU, and GPS measurements, tying in the drone's home position to construct a world-referenced coordinate system.

The YOLOv8 model initiates the detection process, providing base-level object localization, where subsequent algorithms refine the UAV's target approach utilizing height and depth estimations. The convolutional neural network, ActivePerceptionNet, further enhances detection operations through proactive inference, mitigating detection uncertainties from rotating structures like wind turbine propellers. This accounts for periodic downturns in confidence scores typically impacting detection performance.

Key Numerical Outcomes

Simulation results underscored the system's capacity for accurate height and depth estimations with low computational burden. Specifically, the authors report an average height estimation error of approximately 0.06 meters for wind turbines and 0.89 meters for electric towers. Highlighting the increased detection confidence level, nearly peaking the model's capacity, their ActivePerceptionNet integration proved effective, especially within the periodic motion context of wind turbines.

Implications and Future Directions

The implications for UAV active perception systems expand across civil and commercial applications, where cost-efficient and reliable navigation solutions are in heightened demand. The integration of low-cost sensors with advanced perception and control algorithms demonstrates feasible applications in photorealistic simulation environments with potential adoption in real-world scenarios.

Future work could explore real-life validations, extending the algorithm's adaptability across various UAV setups and environmental conditions. The exploration of further sensor types and integration methodologies may enrich the UAVs' perceptual robustness, enhancing overall operational efficacy across diverse operational fields.

In summary, the presented paper offers a perceptive evolution for UAV navigation systems by bridging the gap between advanced computer vision and robust control mechanics using economically accessible sensor frameworks. This work poses a promising avenue for engendering scalable and efficient UAV systems for complex navigational tasks in diverse fields of operation.