Abstract: Spacecraft typically lack access to the adversarial orbital threat's game strategy or model parameters and can only rely on historical perception information to infer the current spatial situation, thereby formulating and executing corresponding evasion strategies. To address such incomplete-information orbital threat avoidance problems, this paper constructs a comprehensive perception-decision game indicator by integrating the orbital environment and aiming to limit the adversarial threat’s perception capability. Based on this, an event-triggered impulsive evasion control problem model is established, and an intelligent evasion control method based on Deep Q-Network (DQN) is proposed. By integrating the spacecraft's impulsive control characteristics with threat perception information, event-triggering conditions are designed to enable the spacecraft to autonomously determine the timing for threat countermeasures. The event-triggered mechanism is combined with a deep reinforcement learning algorithm to design a DQN-based intelligent evasion control law. Furthermore, a reward function that comprehensively considers safety, fuel consumption, and lighting conditions is constructed to enhance evasion capability when the adversary’s game strategy is unknown. The effectiveness and fuel-saving advantages of the proposed method are verified through simulation examples.