Trajectory Planning Algorithm For Uninhabited Aerial Vehicle

This thesis presents several trajectory optimization algorithms for uninhabited aerial vehicle (UAV) and constructs a set of intelligent class library facilitating flexible flying, control and simulation. Path planning has been divided into two hierarchical structures:global path planning and local path updated. In the large scale global path planning, terrain factors, such as mountain range, have been employed to dodge adverse objects. Threatens are measured by 2D probability density; terrain data get from digital elevation map, then a multiple objective function, integrating measurable threatens, average global path altitude and total distance, is constructed. Genetic algorithm is used to solve this multiple objective optimization. In evolution computing, relative distance-angle coding strategy is employed, which naturally involves the physical constrains, such as maximum turning angle and minimum straight line step, into coding variables. Then fitness function is designed based on multi-object fuzzy satisfaction function. And roulette wheel selection, two points crossover and Gaussian mutation strategies is applied, the optimization results are stable. In particular, to improve computational efficiency, a series of methods is adopted:locating neighbor points by binary search, terrain height interpolation by isoparametric element, line integration by distributed cumulating, and group evolution by multi-process. In the local path planning, stricter obstacle avoidance judgment has been employed. In the purpose of real-time optimization output, we remodel the previous model with linear programming (LP), and write a script to translate the mathematical formal to the algorithmic formal so as to take advantage of existing powerful LP solver, such as CPLEX and GLPSOL. Finally, we restructure the output data format to integrate with MATLAB for visualization.In addition, a prototype of UAV intelligent flying facilities has been designed aimed at enhancing the comprehensive UAV's path planning, navigation and visualization. These include UAV path planning related class library, dynamic model and geographic information database. At the viewpoint of practical application, great efforts have been taken to improve the algorithmic robustness in aspect of waypoint visibility, obstacle avoidance and trajectory generation, and the negative effect causing by divisor or parallel coordinate has been removed. Also, a function for full freedom UAV dynamic motion has been written by programming language C, which takes thrust, elevator, aileron, and rudder deflection as its input, and 12 derivatives of the full freedom variables as its output, as well as retaining aerodynamic force interface. This function could be compiled under multiple platforms and linked with other class library for UAV dynamic flying simulation. Finally, in purpose of supporting UAVs' request for threatens, targets or airports information around them in real-time during a flight, an initial geographic information database has been designed on global scalable infrastructure App Engine, which is used to store the potentially huge dataset of global geo-information needed by UAV. What's more, a novel query strategy has been taken to search the adjacent region, which only implements once equation filter constrain and retrieve all entities in a specified adjacent rectangle region. The adjacent region is detected by GPS, and the LatLon data communicate to our database by Map API. This method is particularly meaningful for query on the very database constructed on BigTable architecture, which inherently do not support multiple inequality filters on different properties, such as latitude and longitude, simultaneously in a query.All algorithms in this thesis designed for UAV's path planning and auxiliary has been realized on computer, and their efficiency has been demonstrated by simulation. The real computation employs a network enabled remote central optimization but distributed simulation mode, which makes a sense for constructing large scale sky-ground cycle with human in loop for UAV flying.