In complex interference environments, factors such as noise, occlusion, or interference cause outliers to appear in sensor measurements. These outliers lead to a reduction in the accuracy of the robot’s motion state estimates and can even cause divergence in the Kalman filter. This paper proposes an outlier mitigation method based on geometric constraints and dynamics constraints. The method constructs robot’s geometric constraint equations and dynamics constraint equations to detect outliers in sensor measurements. Additionally, this paper optimizes the step of the unscented Kalman filter to mitigate the detected outliers in real time. Simulation experiments demonstrate that the proposed method effectively combines geometric constraints and dynamics constraints to detect and mitigate outliers in sensor measurements, significantly improving the accuracy of robot’s motion state estimation and showing great application potential.