Abstract: Existing shearer positioning methods suffer from cumulative inertial navigation error drift under complex underground operating conditions, and fixed map constraints have problems such as local map inaccuracy and abrupt constraint changes at map boundaries. A shearer inertial navigation positioning method based on robust adaptive map constraints was proposed. The non-holonomic constraint and forward velocity constraint during shearer movement were fused to construct a velocity measurement model. The discrete point set of the scraper conveyor was reconstructed using hydraulic support heading angle and pushing stroke information, and a map constraint measurement equation was established. An adaptive robust mechanism was introduced into the Kalman filtering framework, and innovation adaptive estimation was used to adjust the map measurement noise covariance online to realize adaptive allocation of map constraint weights. A fade-out map gradual withdrawal mechanism was introduced so that map constraints were smoothly weakened and withdrawn from updates in boundary sections or local failure areas, avoiding filtering oscillation in boundary or failure areas. Experimental results showed that the method maintained stable positioning performance during three consecutive cuts. Under conditions with partial map errors, compared with the fixed map constraint method, it reduced the Root Mean Square Error (RMSE) in the advancing direction from 0.816 4 m to 0.084 7 m and the Circular Error Probable (CEP) from 0.474 6 m to 0.132 5 m, indicating that innovation adaptive estimation dynamically adjusted constraint weights according to changes in map credibility and reduced the interference of inaccurate maps on positioning results. Under conditions with partial map loss, compared with the method without the fade-out map gradual withdrawal mechanism, it reduced the advancing-direction RMSE from 0.550 1 m to 0.154 8 m and the CEP from 0.354 1 m to 0.149 0 m, indicating that the fade-out map gradual withdrawal mechanism improved the smoothness and trajectory continuity during transition stage of map constraint failure.