Vision-based pose estimation for cooperative navigation of AUVs

Determining the position and attitude of AUVs in formation through mutual communication and visual observationamong AUVs will be of considerable value in future multi-AUV cooperative missions. In this paper, an optical navigation pose decoupling model is proposed, which can select the angle between the line-of-sight (Los) vectors of thecoordinated AUVs as the observation value when it is difficult to obtain the accurate distance information betweentheAUVs. In the context of AUV formation communication, the pixel observation equations allow for the decouplingofposition and attitude, enabling highly accurate and computationally efficient solutions. Then, the observabilityof thecoordinated AUVs is evaluated by the observation matrix of the observation angles, and the effects of different formation configurations on the navigation accuracy are analyzed. Based on the analysis, the observation angleofcooperative AUV is optimally selected based on information entropy to improve the position estimation accuracy. Finally, this paper concludes by testing the proposed navigation algorithm via Monte Carlo simulations and provides ananalysis of the impact of different factors on navigation performance.