INS/Vision Integrated Navigation Algorithms Assisted By Vehicle Motion Characteristic Constraints

With the development of science and technology,navigation and positioning technology is playing an increasingly important role in daily life,even in military or aviation applications.Due to the maturity of satellite navigation system,the Global Position System(GPS)-based navigation and positioning system has gradually become an indispensable part of the navigation and positioning field.In order to obtain a more stable navigation system,the integrated navigation system has become the focus of research in this field.The satellite navigation system can provide long-term and high-precision navigation information,and the inertial navigation system(INS)can provide short-term high-precision navigation information without interference from the external environment.So the integrated navigation system based on GPS/INS has become one of the most widely used integrated navigation technologies.However,when the carrier is in the environment where the satellite signals are disturbed or even shielded,such as buildings,forests,underwater or tunnels,the satellite navigation system can not provide reliable navigation information.At the same time,the inertial navigation system has the disadvantage of error accumulation,which makes it unable to provide effective positioning information for a long time.As a result,the INS/GPS integrated navigation system can not work properly.Thanks to the continuous maturity of image processing technology,vision systems are increasingly applied in various fields,especially in the recent emerging fields,such as autopilot,unmanned aerial vehicle,robots.Firstly,this thesis introduces the basic model and principle of visual navigation system,including several major coordinate systems and camera models.In the visual navigation system,feature extraction and matching is a very important step.In this thesis,the commonly used feature extraction and matching algorithms are briefly introduced.Considering the advantages and disadvantages of each algorithm,ORB feature extraction and matching algorithm is adopted and introduced in detail.At the same time,the visual odometer based on ORB algorithm is studied.The feasibility of the algorithm is verified by experiments.Aiming at the problem of inertial navigation error accumulation,the vehicle motion constraint model is analyzed at first.In the process of vehicle motion,if it is assumed that the vehicle will not jump off the ground or drift laterally under ideal condition,the motion constraint conditions can be obtained.By introducing the motion constraint conditions into the solution process of inertial navigation system,the inertial navigation system can be assisted to reduce the error accumulation in the positioning process.Finally,the effectiveness of the algorithm is verified by simulation experiments.Extended Kalman Filter(EKF)is commonly used to estimate system errors and compensate the results.After estimating the error by EKF,there are still many residual high-order error terms in the system results.These errors are mainly caused by non-linear higher-order errors of system error,which are often ignored in the EKF models.Therefore,this thesis proposes a method combining recursive fast orthogonal search algorithm with EKF.The experimental results show that the proposed algorithm is effective.Finally,an INS/Visual integrated navigation algorithm is proposed.The vehicle motion characteristic constraints and recursive fast orthogonal search algorithm are introduced to reduce the system errors and improve the navigation and positioning accuracy.Experiments show that the algorithm can still provide reliable navigation information when satellite signal is affected.