This paper presents a novel whole-body humanoid robot, "Adam," and a novel imitation learning framework based on adversarial motion priors. The authors address the challenges of designing complex reward functions and training sophisticated systems in humanoid robotics. Adam, a full-size humanoid robot with a cost-effective and modular design, is equipped with high-performance actuators and a biomimetic torso configuration, enabling it to mimic human-like locomotion. The proposed framework uses human motion data to guide the learning process, significantly improving the robot's adaptability and performance in complex locomotion tasks. Experimental results demonstrate that Adam can achieve human-comparable performance in various locomotion scenarios, marking a significant advancement in humanoid robotics. The framework's effectiveness is validated through cross-validation experiments using both simulation and real-world testing, highlighting its potential for practical applications in humanoid robotics.This paper presents a novel whole-body humanoid robot, "Adam," and a novel imitation learning framework based on adversarial motion priors. The authors address the challenges of designing complex reward functions and training sophisticated systems in humanoid robotics. Adam, a full-size humanoid robot with a cost-effective and modular design, is equipped with high-performance actuators and a biomimetic torso configuration, enabling it to mimic human-like locomotion. The proposed framework uses human motion data to guide the learning process, significantly improving the robot's adaptability and performance in complex locomotion tasks. Experimental results demonstrate that Adam can achieve human-comparable performance in various locomotion scenarios, marking a significant advancement in humanoid robotics. The framework's effectiveness is validated through cross-validation experiments using both simulation and real-world testing, highlighting its potential for practical applications in humanoid robotics.