| Visual-Inertial navigation is a key technology for UAV autonomous flight in unknown environments.Currently,Visual-Inertial Odometry(VIO)can only solve the navigation requirements of a single UAV.However,for a multi-UAV system,If the equipment such as visual inertial sensor and airborne computer is equipped,the hardware cost will be very high.In view of this,on the basis of traditional VIO,VIO is combined with path decision making,aiming at minimum cost,and a low-cost navigation method for UAV based on visual-inertia fusion is proposed in this paper.The overall process is divided into two stages: the first stage is single aircraft exploration in unknown environment,and the second stage is flight with minimum energy consumption under known map conditions.The main research work is as follows:(1)In stage 1,this paper proposes the Self-synchronizing Monocular VisualInertial Odometry(SMVIO)to solve the problem that the low-cost sensor assembly is prone to time-stamp misalignment.Firstly,the observation error models of camera and IMU were established,and the visual,IMU and marginal residual terms were constructed respectively.The visual residuals were improved,the influence model of time delay and pixel movement was established,and the time correlation between camera and IMU was strengthened.Finally,the real-time correction of camera timestamp and the accurate estimation of pose are realized by nonlinear optimization algorithm.On the other hand,in view of the weak density and low practicability of the current monophthalmic SLAM system,a 3D Mesh raster map reconstruction algorithm is proposed in this paper.Firstly,a two-dimensional Delaunay triangulation algorithm based on point-by-point insertion method was used to construct a local threedimensional Mesh map.Secondly,a two-sided Laplacian smoothing algorithm based on dynamic coefficients is proposed to effectively eliminate and optimize the abnormal structure in the local Mesh map.Finally,a 3D raster probabilistic map model is established,and a probabilistic updating method of local Mesh map is defined to reconstruct a 3D Mesh raster map with more descriptive and dense environment in real time.(2)In the second stage,in order to realize the low-cost flight of the subsequent UAV,this paper takes the three-dimensional Mesh grid map as the input and proposes A three-dimensional A* algorithm based on the minimum energy loss.Firstly,the node representation method and search strategy of the traditional 3D A* algorithm are studied,and the path search time and iteration times are reduced by bidirectional alternate search.On the other hand,considering the 6DOF motion limitation of UAV,the feasible connected domain of the current path node is screened to effectively constrain the motion direction and minimum turning radius of UAV.Finally,the energy loss model is defined and combined with the inspiration function to realize the path calculation of the minimum energy consumption of UAV. |
