AIS-based Vessel Trajectory Prediction for ASV Collision Avoidance
MetadataShow full item record
The automotive industry has already taken a big step towards fully autonomous vehicles.This trend is now spreading to the maritime industry with development ofautonomous surface vessels (ASVs). A reliable collision avoidance (COLAV) systemis essential in this context. In order to avoid collision with other vessels, one mustpredict the future trajectories of nearby vessels. The constant velocity model (CVM)is the prevailing approach for trajectory prediction in today s COLAV systems. Themain purpose of this thesis is to investigate to what extent it is possible to predict futurevessel trajectories based on historical automatic identification system (AIS) datafor prediction horizons up to about 15 minutes. A survey of other relevant predictionmethods are also presented. Two new, AIS based methods for vessel trajectory prediction are developed andtested: the single point neighbor search (SPNS) method and the neighbor coursedistribution method (NCDM). Three speed prediction methods are also tested: thestraightforward constant speed method, a method using the median speed of the predictedstate s close neighbors (CNs) and lastly a linear transition in predicted timebetween the two former methods. The SPNS method is compared to a CVM approach and yields significantly betterresults on curved trajectories, in terms of lower average and median path andtrajectory errors. However, a major part of vessels transit time is spent on straightline trajectories. The SPNS algorithm shows also good path predicting capabilities onclose to straight line trajectories, although the CVM method yields the lowest errors insuch environments. The SPNS algorithm outputs a single predicted trajectory whichtends to follow the most AIS-dense sea lane ahead. Hence, it does neither facilitateany uncertainty measure nor the possibility to suggest multiple possible route choices.The more computational demanding NCDM algorithm, which outputs multiple predictedtrajectories, does better facilitate prediction uncertainty and it is also capableof dividing the predicted trajectories into multiple branching sea lanes. Its predictedpositions at certain time instants are clustered with the density-based spatial clusteringof applications with noise (DBSCAN) algorithm. The predictions are furtherstatistically evaluated with respect to the distances to the nearest cluster centers. Lastly, a computation time analysis is presented. The computational time is reducedfrom an earlier version of the algorithm by storing data in a k-d tree. However,the NCDM algorithm with the tested decision parameters is not practically feasible inreal-time. Several suggestions to reduce the computation time are given.