An Improved Adaptive Sampling Algorithm
Wu Lian, Yiping Li, Shuxue Yan, Jian Liu · 2018
For the observation of the Harmful Algal Blooms (HABs) whose chlorophyll concentration obeys the Gauss distribution, an improved adaptive sampling method for a small autonomous underwater vehicle (AUV) based on the Gauss Process Regression (GPR) is proposed. The adaptive sampling process is divided into three stages: the search, the comb sampling and the escape of the hotspot region. This method enables AUV to switch the sampling stages, online path planning and complete the rapid observation of the unknown area through updating the sampling information of its own environment. Under the environment of four hotspot regions, the comparison simulation experiment proves the effectiveness of the algorithm, which can quickly observe the interest regions and obtain the low error estimation of the characteristic distribution. It also proves that the algorithm can effectively avoid the repeated sampling to the same hotspot region, further improve the observation precision of the hotspot region and reduce the prediction error of the hotspot region.