本研究整合姿態感測模組、救援氣囊與可掛載之概念，開發AUV自主式水下載具掛載式智能救援系統，當智能救援系統判定AUV失能、墜落或嚴重碰撞時，掛載於AUV的智能救援系統可立即啟動CO2充氣機構對尼龍氣囊充氣，能即時對氣囊充氣產生向上浮力，以降低自主式水下載具失誤機率與搜救困難度。其中姿態辨識演算法則是先定義水下姿態之特徵值，並對水下載具的姿態矩陣擷取、降低維度，以提高倒傳遞類神經網路對AUV運動姿態的辨識率。

This research is based on the attitude sensing algorithm to design a portable intelligent rescue system for AUVs. To lower the possibility of losing the underwater vehicle and reduce the difficulty of rescuing, when an AUV intelligent rescue system (AIRS) detects that the fault of AUVs and could not be reclaimed, AIRS can pump carbon dioxide into the airbag immediately to make the vehicle resurface. AIRS consists of attitude sensing module, double-trigger inflator mechanism, and activity recog-nition algorithm. The sensing module is an 11-DOF sensor that makes up of a six-axis inertial sensor, a three-axis magnetometer, a barometer, and a thermometer. Furthermore, the signal calibration and extended Kalman filter (SC-EKF) is proposed to be used subsequently to calibrate and fuse the data from the sensing module. Then, classify the attitude data with the principle of feature extraction (FE), and backpropagation network (BPN) classifier. Finally, the designed double-trigger inflator not only can trigger by electricity but also trigger by water damage when the waterproof cabin is severely broken. With the AIRS tech-nology, the safety of detecting and investigating of using AUVs can be increased since there is no need to send divers to engage in the rescuing mission under the water.