Abstract
Unmanned surface vehicles (USVs) can encounter undetected moving obstacles while sailing along a planned global path. USVs need to plan collision avoidance trajectories for moving obstacles. In this paper, an algorithm based on the Gaussian mixture model (GMM) and Gaussian process regression (GPR) is proposed to predict the motion of moving obstacles and estimate the uncertainty of the prediction. A nonlinear finite-time velocity obstacle (NLFVO) method is developed for obstacle avoidance. The NLFVO method analyzes the velocity of the USV and the predicted uncertain velocity vectors of the moving obstacles and selects a collision-free velocity for the USV and minimizes the objective function. To enable the actual navigation of USVs, the International Regulations for Preventing Collisions at Sea (COLREGs) are considered in addition to the NLFVO method. The simulation results show that the prediction algorithm can effectively predict the trajectory of moving obstacles, and the NLFVO method can obtain a collision-free trajectory for a USV.
Highlights
An unmanned surface vehicle (USV), as a kind of autonomous marine vehicle, is suitable for tasks that are dangerous or unsuitable for manned vehicles
This paper proposes a trajectory prediction algorithm based on Gaussian mixture model (GMM)-Gaussian process regression (GPR) and a USV collision avoidance algorithm based on a nonlinear finite-time velocity obstacle (NLFVO) method
(2) According to the motion uncertainty of the moving obstacle, a collision avoidance trajectory planning algorithm based on the NLFVO method and the dynamic window approach (DWA) is proposed for collision avoidance considering the changes in the velocities of obstacles, collision time, and kinematic constraints
Summary
An unmanned surface vehicle (USV), as a kind of autonomous marine vehicle, is suitable for tasks that are dangerous or unsuitable for manned vehicles. This paper proposes a trajectory prediction algorithm based on GMM-GPR and a USV collision avoidance algorithm based on a nonlinear finite-time velocity obstacle (NLFVO) method. (2) According to the motion uncertainty of the moving obstacle, a collision avoidance trajectory planning algorithm based on the NLFVO method and the DWA is proposed for collision avoidance considering the changes in the velocities of obstacles, collision time, and kinematic constraints. Puo 2 where Puo = Pusv − Pmo. Geometrically, if (9) and (12) are true, that is, the USV will collide with the moving obstacle in the future, the angle χuo of the velocity vector vuo is within the angle between the left tangent λL and the right tangent λR of D(P, rsum) emitted from Pusv.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call