Abstract: To address the problems of accumulated drift in inertial navigation, local mismatch of fixed map constraints, and abrupt constraint variations at map boundaries in shearer positioning systems under complex underground working conditions, a robust adaptive map-constrained INS/odometer integrated navigation method is proposed. Non-holonomic constraints and forward velocity constraints during shearer motion are incorporated to construct the velocity measurement model of the integrated navigation system. The discrete point set of the scraper conveyor is reconstructed from the heading angles and advancing strokes of hydraulic supports, on the basis of which the scraper-curve map model and the map-constrained measurement equation are established. Innovation-based adaptive estimation is employed to online adjust the covariance of map measurement noise, enabling adaptive allocation of map-constraint weights. In addition, a fade-out mechanism is introduced to ensure that map constraints are smoothly weakened and withdrawn from the update process in boundary sections or locally invalid regions. Comparative experiments involving three consecutive cutting cycles were carried out on an intelligent fully mechanized mining experimental platform, and the proposed method was compared with fixed map-constrained and adaptive map-constrained methods. The results show that the proposed method maintains relatively stable positioning accuracy over the three consecutive cutting cycles. Compared with the method without adaptive map constraints, the Y-axis RMSE decreases from 0.8164 m to 0.0847 m, corresponding to a reduction of 89.6%, while the CEP decreases from 0.4746 m to 0.1325 m, corresponding to a reduction of 72.1%. These results indicate that the proposed method can effectively reduce the accumulated inertial navigation error of the shearer and improve positioning continuity and stability under conditions of map boundary transitions and local map mismatch.