Design And Research On Indoor Autonomous Aircraft For Investigation Based On Vision

In recent years,the unmanned aerial vehicle(UAV)has caught widely attentions.The quadrotor has become the research hotspot in the UAV area as its simple structure,excellent maneuverability,abilities of hovering,takeoff and landing vertically and other characteristics.What's more,it has obviously advantages when monitoring,detecting and rescuing.As we all know,the widely used global positioning system(GPS)is not suitable for indoors,tunnel and other similar environments;the motion capture system just can be used in the laboratory and other specific environment.As to solve the need of investigating the dangerous indoors environments task which is not suitable for people to execute such as chemical leaks,biological contamination and etc.The paper has designed an autonomous detection quadrotor that combines a vision,optical flow,inertial sensor and other sensors.First,the whole system is analyzed and divided into five parts: the body,flight control system,visual navigation system,ground station system and remote controller.It has described the composition of the hardware on detail,designed a visual navigation system and flight control system based on ROS and Nutt X,respectively.Second,the paper has designed the flight control system's pose controller and attitude estimator.Having built the non-linearization model according to the newton-euler model.Having designed and simulated the cascade PID pose controller of the quadrotor and the attitude estimator based on the complementary filtering principle.Then,we have analyzed the visual simultaneous localization and mapping(SLAM)algorithm.Having introduced the model of camera and the method for extracting and matching the feature point,analyzed the method of restoring camera position between two frames combining the epipolar geometry principle from that.The article has designed to compose the map based on feature points and keyframes and optimized the visual odometer by using the bundle adjustment.Implementing the whole visual SLAM founction through analyzing the loop closing detection problem and method.Next,we have designed navigation algorithm that combined many kinds of sensors on the quadrotor.According to the problem of monocular visual scale,adopting the ultrasonic sensor and the maximum likelihood estimation to calculate monocular visual scale.According to cascaded PID position control's requirements for the information of position and velocity,adopting the extended kalman filter(EKF)fusions vision,inertial measurement unit(IMU),barometer and other sensors to estimate the position and speed.According to lack of environmental texture,using the visual SLAM and optical flow to satisfy the normal flying.According to the need of navigation map,h aving implemented a semi-dense map and Octo Map.According to the need of autonomous investigation in unknown environment,having designed the method of stepping self-exploration.Finally,the design of the aircraft has been tested in real environment.The results show that the aircraft can realize the autonomous navigation in the unknown indoor environment and complete the autonomic investigation mission.