Autonomous robots are widely used in modern industries and daily life. The global navigation satellite system (GNSS) real-time kinematic (RTK) can provide centimeter-level positioning accuracy in open areas but is significantly challenged in urban canyons, due to the severe signal reflections. To fill this gap, this paper proposed a tightly coupled (TC) GNSS-RTK, inertial navigation system (INS), and odometer (RIO) integration via factor graph optimization (FGO) aided by the GNSS outlier mitigation (see Fig. 1). Different from the existing EKF, the FGO enables the explorations of the correlation within the historical measurements, which leads to improved robustness towards the unexpected outlier measurements. One of the challenges is that the direct integration of the high-frequency INS and odometer measurements can lead to unacceptable computational load using the factor graph model. To cope with the high-frequency raw INS and odometer measurements, the pre-integration is adopted to integrate the high-frequency measurements to maintain computational efficiency. The relevant derivations of the pre-integration model are rigorously formulated together with the Jacobian matrices. Moreover, the marginalization-based sliding window is adopted to achieve real-time performance. To further mitigate the impacts of the potential GNSS outlier measurements, the measurement quality management strategy is proposed by combining the robust estimation and odometer/INS-aided consistency check. Thanks to the accurate relative pose change from IMU and Odometer pre-integration, the system could detect and exclude GNSS outliers in both pseudo-range and carrier-phase measurements, after which ambiguity resolution (AR) is performed with high-quality measurements. To validate the effectiveness of the proposed method, several real-world datasets are collected to validate the proposed method in GNSS-challenging environments. Our experiments showed that our method achieved higher accuracy and consistency than state-of-the-art methods (See Fig. 2). To the best of the author’s knowledge, this is the first work proposing the tightly coupled GNSS-RTK/INS/Odometer integration using FGO.