<p>This paper proposes a novel method for optimizing the whole-body walking pattern of humanoid robots by integrating pelvic yaw rotation, heel–toe motion, and arm swing. These motions, which mimic human walking, are optimized in real-time using a Hierarchical Quadratic Programming (HQP) approach based on task priority order. The framework leverages the redundancy of the lower body to optimize pelvic yaw rotation and the redundancy of the swing foot for heel–toe walking. In addition, it introduces a method to minimize the distance between the swing leg and support leg to enhance toe-off performance. The proposed method is rigorously evaluated for its impact on robotic walking, considering various aspects such as walking patterns, step length, knee joint angular velocity, potential for singularity occurrence, and computation time. Comparative analysis is performed through simulations to assess the combined effects of pelvic rotation and heel–toe walking, examining parameters like maximum step length, knee joint dynamics, occurrence of singularities, and energy consumption. The experimental results confirm the effectiveness of the proposed approach, demonstrating improvements in step length and overall walking efficiency. This work advances the field by addressing the simultaneous incorporation of pelvic rotation and heel–toe walking in humanoid robots, a combination that has not been extensively explored in prior research.</p>