Autonomous Driving Systems (ADSs) are safety-critical systems, where even minor errors can lead to significant consequences. Therefore, thorough testing of ADSs is essential. Scenario-based testing is a widely adopted approach for evaluating ADS performance. By executing the ADS within specifically designed scenarios, engineers can identify and address system errors to enhance safety. However, efficiently generating critical scenarios remains a challenge. In this work, we propose a novel and efficient method for generating critical scenarios by minimizing temporal loss in space-sharing conflicts to guide the fuzzing process. We evaluate our method using the CARLA simulator with Autopilot as the target ADS. The results demonstrate that, compared to two widely used methods, the proposed approach effectively and efficiently generates a diverse set of critical scenarios.