Objective:To evaluate the robustness of fully automated adaptive planning for Ethos online adaptive radiotherapy (ART) based on the intelligent optimization engine (IOE).Methods:Clinical data of 11 stage ⅠB cervical cancer patients admitted to Peking Union Medical College Hospital between June 2021 and June 2022 were retrospectively analyzed. Original planning images and iterative cone-beam computed tomography (iCBCT) images of each radiotherapy treatment were acquired, and all patient data were imported into the Ethos simulator. IOE-based 9-field automatic plan generation was performed for 11 patients using Ethos, and the generated plans were sent to online adaptive radiotherapy simulation to obtain each online adaptive radiotherapy plan (273 fractions in total) and complete the simulated treatment. For comparison, manual plan design was performed based on the images and contoured structures used for online adaptive radiotherapy planning, and the manually plans created with evenly divided 9 fields. Dosimetric parameters, plan complexity parameters, and Mobius quality assurance (QA) pass rates were collected to compare and evaluate the robustness of the online adaptive radiotherapy plan in terms of organs at risk (OAR), target volume dosimetric parameters, and plan complexity by using paired t-test or rank sum test. Results:The online adaptive plan of cervical cancer had comparable planning target volume (PTV) coverage compared to the manual plan. For the clinical target volume (CTV) D 99%, online adaptive plan was significantly higher than the manual plan [(45.93±0.36) vs. (45.32±0.31) Gy, P<0.001]. For hot dose area, the maximum point dose (PTV D max) of adaptive plan was significantly higher than the manual plan [(49.89±1.25) vs. (48.48±0.77) Gy, P<0.001], but the PTV D 1% of adaptive plan was significantly lower than the manual plan [(47.22±0.29) vs. (47.59±0.48) Gy, P<0.001]. There was no statistical difference in the conformal index ( P=0.967). And there was significant difference in the homogeneity index, with same medians and less dispersion in adaptive plan ( P<0.001). For OAR dose, bladder D mean, rectal V 40 Gy, small intestine D mean of adaptive plan was slightly higher than that of the manual plan; the rectal D mean, small intestine D 2 cm3 of the adaptive plan was slightly lower than that of manual plan; dosimetric parameters of right and left femoral heads, spinal cord and bone marrow of the adaptive plan were better than those of manual plan. The adaptive plan had more monitor units (MU) than the manual plan, but the complexity of the adaptive plan was significantly lower than that of the manual plan (0.135±0.012 vs. 0.151±0.015, P<0.001). For Mobius γ pass rate (5%/3 mm), both adaptive and manual plans met clinical requirements. Conclusion:Ethos cervical cancer online adaptive plan, which is based on the IOE engine, demonstrates good robustness and ensures the quality of online adaptive plans generated for each treatment fraction.