Research On Visual-inertial Navigation Method For Cooperative Mars Surface Exploration

Autonomous navigation is an essential technology for Mars exploration.For Mars surface exploration,due to the lack of global navigation coordinates provided by navigation satellites and prior map information,Mars Rovers often use visual navigation methods to navigate in its own local coordinate system.In 2015,NASA proposed to use the Mars Helicopter to explore the surface of Mars.The Mars helicopter is expected to become the “eye in the sky” and effectively expand the exploration area and improve efficiency.However,the thin atmosphere on Mars has severely limited the take-off weight of the Mars helicopter.This means that Mars helicopters cannot carry weight and power-consumption sensors.Compared to the Mars Rover,Mars helicopters need a more lightweight state estimation system.In addition,the yearning for collaborative exploration is undoubtedly a major driving force for the birth of Mars helicopters.Unifying the navigation coordinates of each agent to achieve global mission and path planning is an essential technology for achieving efficient collaboration.In recent years,the measurement accuracy of low-cost IMUs has gradually increased,and visual-inertial integrated navigation has become a research hotspot.By integrating visual and inertial measurement information,state estimation can achieve higher accuracy and robustness.The monocular visual-inertial system consisting of a monocular camera and IMU has advantages in terms of weight,size,and power consumption compared to stereo cameras and lidar sensors.It is ideally suited for flight platforms with limited carrying capacity.Along with the development of computer vision technology,the accuracy of similar image recognition is being improved,which provides conditions for using visual information for place recognition,relocalization,and merging of visual maps.This allows multiple Mars helicopters to work together by processing visual images and unify their navigation coordinate system respectively.This paper aims at the future of Mars surface exploration activities,and studies the visual/inertial integrated navigation method and discusses the feasibility of applying it to multi-Mars helicopters for aerial collaborative exploration tasks.From the theoretical research to the experimental verification sequence,this paper first studies the basic theory of visual/inertial integrated navigation;then it studies the two cores of the integrated navigation method for collaborative exploration,which are sliding window-based monocular visual-inertial odometry,and loop detection,map merging,global optimization methods based on visual feature points.Finally,the performance of the integrated navigation method for collaborative exploration was verified and analyzed through dataset testing,physical experiments and simulation scenario experiments.