基于多传感器信息融合与外形识别的四旋翼飞行器避障算法研究

发布时间：2018-06-24 01:56

本文选题：四旋翼飞行器 + 多传感器信息融合　；参考：《江西理工大学》2017年硕士论文

【摘要】：随着四旋翼飞行器在地质勘探、农业生产、物流运输、抢险搜救等应用领域的不断深入,人们对四旋翼飞行器的要求不断提高,达到自主控制飞行是其发展的必然趋势,而四旋翼飞行器的自主避障则是实现其完全自主飞行的基础。此外,四旋翼飞行器应用的普及化也带来了许多安全问题,各种撞机、坠机事故时有发生,其对人身财产安全带来了极大的威胁。因此展开对四旋翼飞行器航行中及时有效地避开障碍物以保护人身、机身和建筑物的研究具有重要意义。针对四旋翼飞行器的避障问题,文章研究了基于多传感器信息融合与外形识别的避障算法。全文的主要内容包括:(1)搭建了以STM32F407为主控芯片、MPU6050为惯性导航单元、GPS为定位导航系统的四旋翼飞行器硬件平台;采用Kalman滤波器对激光传感器和超声波传感器的测距信息进行融合,并用遗忘因子b对原算法予以优化,然后用MATLAB中的Simulink模块仿真得到最佳遗忘因子b值,完成了Sage_Husa自适应Kalman滤波器的设计;最后对三种材质的障碍物进行了距离信息融合的实验。(2)开展了关于障碍物外形识别的研究。介绍了基于超声波传感器三点直线排布识别障碍物外形的方法,根据其不足和局限之处,研究了一种新的测量方法——超声波传感器三点120°钝角等腰排布测量法。针对五种基本障碍物外形,用几何法推导了其测量公式,阐述了外形识别的原理,为系统的避障算法做好理论铺垫。(3)基于障碍物距离信息和外形信息,制定了四旋翼飞行器的避障算法:首先,根据距离的融合信息判断四旋翼飞行器的安全状态,以确定自动避障系统的开启与否;其次,自动识别障碍物的外形,根据不同外形制定不同的避障措施;然后,利用PID算法实现四旋翼飞行器的避障动作;最后,进行了障碍物外形识别与避障的实验。多传感器信息融合的实验表明:使用单一传感器测距容易出现采集值不稳定和不精确的问题,而利用Kalman滤波器进行信息融合可以输出稳定的障碍物距离信息,并降低测量误差;而Sage_Husa自适应Kalman滤波器的融合效果在原滤波器基础上进一步得到改善。障碍物外形识别与避障的实验证明该避障算法可以识别出障碍物的外形,并做出相应的避障动作,达到避障目标。
[Abstract]:With the deepening of the four-rotor aircraft in geological exploration, agricultural production, logistics transportation, rescue and other applications, people's requirements for the four-rotor aircraft are constantly raised, and it is the inevitable trend of its development to achieve autonomous flight control. Autonomous obstacle avoidance is the basis of complete autonomous flight of four-rotor-wing aircraft. In addition, the popularization of the application of four-rotor aircraft also brings many safety problems. Various collision and crash accidents occur from time to time, which bring great threat to personal and property safety. Therefore, it is of great significance to study the protection of human body, fuselage and buildings by avoiding obstacles in time and effectively. Aiming at the obstacle avoidance problem of four-rotor aircraft, the obstacle avoidance algorithm based on multi-sensor information fusion and shape recognition is studied in this paper. The main contents of this paper are as follows: (1) the hardware platform of a four-rotor aircraft with STM32F407 as the main control chip, MPU6050 as the inertial navigation unit and GPS as the positioning and navigation system is built, and the ranging information of laser sensor and ultrasonic sensor is fused by Kalman filter. The original algorithm is optimized by forgetting factor b, and then the optimal forgetting factor b is obtained by Simulink module in MATLAB, and the design of SageStack Husa adaptive Kalman filter is completed. Finally, the distance information fusion experiment of three kinds of obstacles is carried out. (2) the research on obstacle shape recognition is carried out. This paper introduces a method of identifying the shape of obstacle based on three point straight line arrangement of ultrasonic sensor. According to its deficiency and limitation, a new measuring method-measuring method of isosceles arrangement with three point 120 掳obtuse angle of ultrasonic sensor is studied. According to five kinds of basic obstacle shape, this paper deduces its measuring formula by geometric method, expounds the principle of shape recognition, and lays a good theoretical foundation for the obstacle avoidance algorithm of the system. (3) based on obstacle distance information and shape information, An obstacle avoidance algorithm for a four-rotor aircraft is developed. Firstly, the safety state of the four-rotor aircraft is judged according to the fusion information of the distance to determine whether the automatic obstacle avoidance system is on or not; secondly, the shape of the obstacle is automatically identified. According to different shapes, different obstacle avoidance measures are developed. Then, the four-rotor aircraft obstacle avoidance is realized by using pid algorithm. Finally, the obstacle shape recognition and obstacle avoidance experiments are carried out. The experiment of multi-sensor information fusion shows that using a single sensor to measure the distance is prone to the problem of unstable and inaccurate acquisition value, while using Kalman filter to fuse the information can output the stable obstacle distance information and reduce the measurement error. The fusion effect of SageHusa adaptive Kalman filter is further improved from the original filter. The experiments of obstacle shape recognition and obstacle avoidance show that the algorithm can recognize the shape of obstacles and make corresponding obstacle avoidance actions to achieve the goal of obstacle avoidance.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：V249;TP212

【参考文献】