<p>An intelligent guidance strategy that integrates optimal guidance and meta-reinforcement learning is proposed to address the multi-constraint diving maneuver guidance problem for hypersonic vehicles. In this context, guidance must be optimized to satisfy constraints on the terminal position and angle of the diving vehicle. To that end, bias commands are introduced to avoid interception through additional longitudinal and lateral maneuvers. Given the high nonlinearity of the control model and the difficulty in accurately predicting time-to-go, the bias commands are determined using a meta-reinforcement learning model. Thus, corresponding state and action spaces are constructed to train the computational agent along with a reward function that comprehensively accounts for energy consumption, terminal constraints, and miss distance. The results of numerical simulations show that the proposed intelligent strategy can execute evasion maneuvers effectively and provide precision guidance. Moreover, it exhibits superior generalization and faster adaptation to new tasks compared to traditional deep reinforcement learning methods.</p>