Research And Implementation Of Indoor Obstacle Avoidance And Navigation Based On Small Size Quad-rotor Helicopter With OS/JVM

Micro air vehicles are widely used to implement complex tasks in terms of investigation, rescue work, vedio monitoring, algritural pest control and geological exploration due to their good performance. Small size quad-rotor helicopter is regarded as a typical micro air cehicle, and it has the advantages of flexible, low-cost, smaller size and convenient and so on, so it can used in the dangerous environments. The researches about indoor flight for small size quad-rotor helicopter have attracted widely attentions nowadays, and these researches are mainly focuses on the obstacle avoidance and navigation and so on.Other universities focuses on the obstacle avoidance and navigation of small size quad-rotor helicopter by adding extra sensors such as RGB cameras, laser finder, ultra wideband, GPS and so on. The main innovation for this thesis is that researches about obstacle avoidance and navigation are all based on the small size quad-rotor helicopter only with single camera, which-is executed on the designed high performance software platform of OS/JVM.Firstly, in order to achieve best performance about real-time images handling and high stability. This thesis designs a light-weight embedded operating system core for the controller of the small size quad-rotor helicopter, and a java virtual machine, then a good performance software platform of OS/JVM is established so as to improve the real-time execution ability of the small size quad-rotor helicopter. The experiments in terms of real-time ability and network performance are tested, besides, the task scheduling algorithm is also improved so as to reduce the overhead of the task and make the operating system kernel have the features of high real-rime ability and low delay.Next, a reduced JAVA virtual machine FijiVM is also implemented for the light- weight operating system kernel. This thesis presents the task execution mechanism and accordingly class constructing detailly. The later pratical experimental results in terms of obstacle avoidance and navigation, show that the software platform of OS/JVM formed by united optimization of operating system kernel and JAVA virtual machine can well be used to serve the real-time JAVA image application, the most important is that it can also meet the real-time ability requirement about images collected from the single camera of the small size quad-rotor helicopter.Next, this thesis implements the obstacle avoidance for the small size quad-rotor helicopter in the indoor environment, and the innovation for this is that the optical flow is introduced to effectively resolve the problem of dynamic obstacle avoidance. About the optical flow principle, the merged optical flow algorithm is proposed based on the practical optical flow of the detected objects and the movement optical flow caused by the movement of the camera in the small size quad-rotor helicopter. The experimental results show that the merged optical flow algorithm can achieve good performance for the obstacle avoidance for the small size quad-rotor helicopter in the indoor environment. Finally, the real-time ability and the accuracy rate are both tested and analyzed in detail. For the autonomous flight of the small size quad-rotor helicopter in the indoor environment, the innovation is that this thesis implemented the vanishing point algorithm used to achieve autonomous navigation. The system obtains all the straight lines in the images after the procedure of pre-processing, Canny edge detection, and random Hough transformation. Then system calculates all the intersection points formed by all the straight lines. The vanishing point fast estimation algorithm VPEE is proposed based on the cluster principle. Finally, the experiments about the real-time ability of VPEE algorithm are tested, and the experimental results show that the VPEE algorithm can effectively used to achieve autonomous navigation for the small size quad-rotor in the indoor environmentFinally, the image entropy is introduced to resolve the obstacle avoidance in the indoor corridor, and it together with the balance optical flow algorithm can effectively implement the avoidance of static obstacle and wall with low texture. Multivariable RBF neural network PID control algorithm is presented to improve the flight stability for the small size quad-rotor helicopter. The effects of multivariable RBF neural network PID control algorithm in terms of roll, yaw, pitch and horizontal, vertical errors are all tested and analyzed, and the experimental results show that the multivariable RBF neural network PID control algorithm can better used to achieve the flight control for the small size quad-rotor helicopter in the indoor environment.In a word, based on the small size quad-rotor helicopter with AVR microcontroller, this thesis designs a light-weight operating system kernel and reduced FijiVM, which are optimized together so as to achieve the best performance for the small size quad-rotor helicopter with the real-time JAVA application in terms of real-time ability, stability and determinacy. Besides, the implemented technologies about image entropy with balance optical flow algorithm, merged optical flow algorithm, vanishing point algorithm, and multivariable RBF neural network PID control algorithm used in the obstacle avoidance and navigation provide strong technical support for the researches about the flight control for the small size quad-rotor helicopter in the indoor environment.