For the dynamic observation of oceanic fronts, a data-driven adaptive front-tracking algorithm for autonomous underwater vehicle (AUV) is proposed based on the model prediction of the ambient temperature data obtained by online sampling. Firstly, a dynamic model of front temperature is established by analyzing the temperature characteristics of fronts and water masses on both sides. Secondly, Gauss process regression (GPR) is used to process the real-time AUV observation data and predicts the current location environment model. Finally, an improved gradient search algorithm is used to plan the sampling path. The simulation results show that the proposed method can achieve continuous tracking down the front. By comparing with other front tracking algorithms, the proposed method can effectively track complex fronts, and acquisition of front area data is more efficient.

No abstract

To improve the performance of Particle swarm optimization (PSO) in online optimization problems, a multi-center PSO algorithm with memory ability was proposed. Main strategies of the proposed algorithm include the initial population optimization based on historical optimal solution and improved chaos mapping and the multi-center collaborative search. To verify online optimization performance, the proposed algorithm is applied to the online modelling process of an underwater vehicle thruster to optimize the modeling parameters. Result proves the superiority of the proposed algorithm in online optimization problem. Keywords Online optimization • PSO • Tent mapping • Multi-center collaborative search • Online modeling • Underwater vehicles