drone-cv-expert

Sensor fusion and SLAM enable autonomous navigation by combining multiple sensors (IMU, camera, GPS) through EKF/UKF filtering and visual-inertial odometry. This Skill guides you through integrating these components with MAVLink protocols and Pixhawk/DJI flight controllers to achieve reliable autonomous flight in GPS-enabled and GPS-denied environments.

Real-time object detection on edge devices uses optimized models like YOLO with ByteTrack for efficient inference on Jetson/Orin hardware. This Skill covers deployment patterns, performance optimization, and integration with your drone's perception pipeline for autonomous target tracking and decision-making.

Yes. GPS-denied navigation uses visual-inertial odometry, SLAM, and camera-based object detection to enable autonomous flight indoors or in denied environments. This Skill provides guidance on implementing these techniques with proper sensor fusion and safety considerations for reliable control.

Multi-sensor integration uses Extended Kalman Filters (EKF) or Unscented Kalman Filters (UKF) to fuse data from cameras, IMUs, and other sensors into a unified state estimate. This Skill covers MAVLink protocol implementation, data synchronization, and real-time constraint management for Pixhawk and DJI platforms.

Autonomous drone safety requires pre-flight validation workflows, real-time constraint adherence, and failure-mode handling in perception pipelines. This Skill addresses safety patterns for reliable autonomous flight, including sensor validation and graceful degradation when vision or fusion systems encounter edge cases.

Yes. This Skill covers path planning algorithms integrated with real-time object detection and sensor fusion to enable collision avoidance and autonomous navigation. It includes guidance on adapting these techniques for both structured and unstructured environments across GPS and GPS-denied scenarios.