Research On Path Planning Of Unmanned Surface Vehicles For Task-points In Shallow Sea

The key to realizing intelligent autonomous navigation of Unmanned Surface Vehicle(USV)is whether it can reasonably plan paths based on shallow sea environment information.Due to the limitation of data acquisition and the influence of random dynamic characteristics of the environment,there are still many unsolved problems,for instance,shallow sea modeling problem with limited data sampling ability,safety problem of USV in shallow sea and multi-USV multiple task-points traversing problem,in path planning of USV in shallow water environment.To address these problems and to provide theoretical basis for USV autonomous navigation in shallow sea environment,this thesis investigates the USV path planning problem in three aspects,namely shallow sea environment characteristics representation,dynamic path planning in shallow sea environment and multiple task-points traversing.The main contributions of this thesis are as follows:In order to solve the problem of shallow sea environment information exaction with limited data acquisition performance,a shallow sea environment characteristics representation strategy,which comprises the module for real-time detection of surface obstacles,the module for modeling underwater obstacles and the module for estimating dynamic characteristics of water-level,is proposed in this thesis.To be more specific,firstly,in the module for real-time detection of surface obstacles,the semantic segmentation model based on the Fully Conventional Networks(FCN)is improved by double-branches decoding operation,and the detection accuracy of surface obstacles under various environmental disturbances is improved.Secondly,in the module for modeling underwater obstacles,limited height data of underwater obstacles of shallow sea areas are fitted by the interpolation basis function so as to improve the accuracy of modeling underwater obstacles.Finally,in the estimation module for longterm water-level dynamic characteristics,the estimation method of water-level dynamic characteristics based on Poisson and Gaussian distribution is proposed to estimate long-term discrete and continuous water-level dynamic characteristics around the meteorological data,such as the benchmark water-level/tide-height curves of rising and falling tide and wave fluctuation characteristics.Simulation results verify that the proposed strategy can extract edge detail features of water surface obstacles,restore the shape of underwater obstacles with high-precision and obtain water-level signal values in line with the changing law of long-term water-level dynamic characteristics.In order to solve the problem of navigation safety in shallow sea environment,a dynamic path planning strategy,which comprises the module for modeling water surface map,the antirollover module,the module for avoiding environmental loads and the anti-grounding module,is proposed in this thesis.To be more specific,firstly,in the module for modeling water surface map,the normalization processing is used to improve the segmentation efficiency of K-means clustering main function,and an expansion factor is introduced to expand the edges of obstacles to ensure that the safety area of USV is not infringed.Secondly,in the antirollover module,a path smoothing method based on B-Spline(BS)is proposed.By virtue of the proposed method,the possibility of rollover is reduced by quickly reducing the yaw rate of collision avoidance path.Thirdly,in the module for avoiding environmental loads,a novel algorithm based on deflection factor is proposed to ensure that the yaw angle will adapt to the environmental loads directions.The algorithm considers avoidance of environmental loads and optimization of path length.Finally,in the anti-grounding module,a path correction algorithm based on Navigable Degree-Varying(NDV)and a real-time collision avoidance algorithm based on Gaussian-Extraction(GE)are proposed.The former can ensure navigation safety of USV in weak wind and wave sea environment by building the dynamic models,which are derived from the relationship between the discrete water level and obstacles.The latter can not only improve the efficiency of smoothing the obstacle edges through the Fuzzy Gaussian Mixture Model(FGMM),but also ensure navigation safety of USVs in strong wind and wave sea areas by extracting real-time collision avoidance path points.The reference data for extracting real-time collision avoidance points are derived from the relationship between continuous water-level dynamic characteristics and obstacles.Taking a shallow sea map in Dalian as basis for conducting simulation students,experimental results verify the effectiveness of the modeling method for water surface map and path planning algorithms in the proposed strategy.In order to solve the problem of cooperative traversing multiple task-points under the multiple constraints,a cooperative traversing path planning strategy,which comprises cooperative assignment module,path traversal module and anti-collision module,is proposed in this thesis.To be more specific,firstly,in the cooperative assignment module,a task-points cooperative assignment method based on priority reward mechanism is proposed.By virtue of the proposed method,while the length of traversal paths and work time are optimized,the special task-points are also guaranteed to be assigned priority.Secondly,in the path traversal module,a turning cost optimization algorithm based on Ellipse Guidance Range-Fast Marching Method(E-FMM)is proposed to ensure that the turning angle is limited,and the turning cost of the traversal paths is further optimized by adjusting the turning radius.Finally,in the anti-collision module,a collision reference path planning algorithm based on TangentCOLREGs(T-COLREGs)and a Surge-Adjusting(SA)guidance law are proposed.The former can predict key positions with reference to the length and speed of a USV,and then quickly obtain the collision avoidance reference paths by inserting tangents.The latter can reduce the approach speed between USVs by adjusting the surge reference values in the process of guiding USV navigation to improve the safety of anti-collision.The Lyapunov stability criterion is used to prove the convergence of the proposed guidance law.In addition,according to the working conditions,the states of USVs and changes in shallow sea environment,task-points re-assignment and path re-planning are carried out to enhance the efficiency of multiple USVs traversing multiple task-points.Taking a multiple task-points scene in a shallow sea map in Dalian as the basis for conducting simulation studies,experimental results verify the effectiveness of the proposed strategy.