Design and Implementation of a Multi-Sensor Fusion Quadrotor System for Indoor Wildlife Inspection

Abstract

Indoor wildlife inspection is difficult for small unmanned aerial vehicles because GPS is unavailable, manual patrols are inefficient, and recognition results must be coupled with safe route execution. This paper presents an embedded quadrotor inspection system that integrates UWB positioning, optical-flow velocity estimation, laser/barometric altitude sensing, IMU attitude measurement, LoRa communication and onboard AI vision. The flight controller is implemented on an STM32F429 platform with a position-velocity cascaded PID controller. A grid-based coverage planner generates inspection routes under no-fly-zone constraints, while a MaixCam Pro module performs animal recognition and counting. Experiments in a 5.0 m by 4.0 m indoor testbed show that the system completes inspection flights within 248.6—261.3 s, maintains path deviation within 2.5—3.0 cm, and reaches an average recognition accuracy of 90.0%. The results demonstrate a deployable low-cost solution for GPS-denied indoor wildlife inspection.