<p>Adaptive multi-sensor fusion navigation in complex cross-scenario environments requires adaptive decision-making strategies that can respond to both external environmental changes and internal information degradation. Existing fusion navigation methods typically rely on fixed fusion rules, predefined noise models, or analytically designed parameter mappings, thereby limiting their ability to capture the nonlinear coupling and temporal variation among heterogeneous navigation sources. To address this challenge, this paper proposes a quantitative belief-guided intelligent decision-making method for adaptive multi-sensor fusion navigation. The proposed method constructs a compact quantitative belief state by jointly modeling exogenous availability belief and endogenous credibility belief, which respectively characterize the quality of environmental geometric constraints and the reliability of internal estimation information. Based on this belief representation, a learning-based decision agent is designed to infer adaptive fusion decisions for multi-sensor fusion. By incorporating temporal belief evolution into the decision process, the proposed method enables the navigation system to dynamically adjust the contribution of different navigation sources under varying cross-scenario conditions. Both simulation and real-world experiments demonstrate that the proposed method can effectively characterize changes in environmental constraints and internal information states, learn robust fusion decision policies, and significantly improve navigation accuracy and robustness in challenging cross-scenario environments.</p>