Abstract: To address the problem of low positioning accuracy and poor stability of electric drive mobile platforms caused by object obstruction in protected horticulture scenes, this study designs a combined positioning method based on ultra-wide band (UWB) and inertial measurement unit (IMU). Additionally, an adaptive robust combination positioning algorithm (ARCPA) based on Kalman filtering is developed to effectively improve positioning accuracy and enhance stability. Firstly, a multi-sensor integrated positioning system combining UWB and IMU is constructed. Secondly, a differential distance-based sliding window method is introduced to precisely detect non line of sight (NLOS) errors in the positioning data of UWB. Furthermore, to address the interference of measurement noise fluctuations in Kalman filtering, a cosine restart function is used to dynamically adjust the fading factor in real-time, thereby enhancing the accuracy and robustness of the filter. Finally, a position dilution of precision (PDOP) weighted robust factor is incorporated to improve the resilience of the positioning system against noise and model disturbances in protected horticulture scenes. In order to verify the effectiveness of proposed integrated positioning, an electric drive mobile platform and real-time positioning system are built, then the simulation and real vehicle positioning experiments are conducted in the typical protected horticulture scenes. The experimental results demonstrate that the root-mean-square errors before and after UWB ranging correction under line-of-sight condition are 119.0 cm and 49.0 cm, respectively, and the ranging accuracy improved by 58.8% after correction. The positioning error of single sensor reaches 36.68 cm, while the error of the proposed integrated positioning method is only 6.63 cm, which is 81.92% higher than the positioning accuracy of single sensor, and the maximum error is reduced by 88.89%. This indicates that the proposed integrated positioning method can significantly reduce the impact of NLOS and geometric error on the positioning accuracy of the electric drive mobile platform for protected horticulture. Moreover, the root-mean-square error and maximum positioning error of ARCPA are 17.70 cm and 44.44 cm, respectively. Compared with the Kalman filtering, unscented Kalman filtering, extended Kalman filtering, particle filtering and adaptive robust Kalman filtering algorithms, the accuracy of the proposed method has been improved by 50.15%, 74.63%, 60.01%, 60.88% and 28.34%, respectively, and the maximum positioning error has been reduced by 66.83%, 73.67%, 69.07%, 71.2% and 50.50%, respectively. The finding can provide theoretical basis and implementation plan for positioning and navigation in protected horticulture environment.