| As the world’s growing population and economy rising,passenger and cargo demand is more and bigger ocean transportation is capillary connection sea economy,in order to meet the increasing transport demand,ship number also inevitable rise,lead to the same area of the sea need to accommodate more sailing ship,the ship collision accident is more frequent whether military ships,or civilian vessels,ship collision accident can cause upset sail plan traffic congestion waters threaten the seaman life property damage and other serious negative impact Ship collision avoidance path planning system can provide safe and reliable navigation route for reference,reduce the driving load and human judgment of seafarers,no one can also be used for planning the autonomous navigation of the ship path chip technology’s rapid development in recent years,making computer core processor frequency more and more high,this provides intelligent navigation of ships with good industrial base,under such a background,this paper constructed a combining static global path planning and local path planning of dynamic collision avoidance system.First of all,this paper introduces the collision avoidance path planning technology for ship the importance of autonomous navigation,path planning technology at home and abroad and comparative analysis of the present situation of study of neural network,this paper introduces the ship collision,what are the process of collision avoidance action should be in what stage of ship navigation system can provide area on the various aspects of the data,on the other hand,from the encounter situation environmental liability limit three aspects analyzes the different situation of ship collision avoidance of responsibility,exist in the aspects of ship collision avoidance are summarizedSecondly,obstruction of static global environment change of state is small,the basic does not need to complete the planning adjustment feature,rasterize the global environment,with static obstruction as taboo table,simplify the problem and then this paper introduces the principle of ant colony algorithm,analyses the general ant colony algorithm for optimization of place,in the traditional ant colony algorithm is derived with the elite of ant colony algorithm,ant colony algorithm based on improved elite,in the grid environment of path planning problem in the optimization performance is more outstanding,through the simulation experiments verify the static global path planning method is feasibleFurthermore,the traditional encounter risk estimation accuracy is low,and easily influenced by artificial factors,using Neural Network instead of traditional method for risk evaluation can promote accuracy due to the traditional BP Neural Network can’t handle the limitations of time-series data,thus can handle a time-series data are derived from the RNN(Recurrent Neural Network)Neural Network,but its gradient can’t normal delivery,perfect after the LSTM(Long Short-Term LSTM)network can process the data of ships that meet in time series.The LSTM network can use the data of ships that meet in time series to train the LSTM network to predict the risk with higher accuracy based on the predicted dataFinally,will focus on static global environment planning a good path,on the basis of grid division,to establish a coordinate system,keep the beginning and end of the path in this section,will use the LSTM network calculates encounter risk of ships used in the fitness function of the particle swarm optimization(PSO)algorithm,the use of this method shows the front path planning again,achieve local area of dynamic collision avoidance,the purpose of using software to build simulation validation of this approach is feasible. |
