Robots are increasingly expected to work alongside humans in unstructured environments and perform complex tasks that involve free and constrained motion. Accurate tracking in free motion and compliance during interaction become essential as control objectives for successful task performance and safety. This study presents a novel control structure that achieves these objectives without relying on force/torque (F/T) sensors or joint torque sensors. The proposed method modifies the nonlinear disturbance observer (DOB) structure to modulate disturbance compensation based on exogenous force monitoring, distinguishing between disturbance and interaction forces using a pre-estimated compensation zone. To ensure stable contact with unknown environments, a compensation modulation method based on passivity analysis is provided. In addition, we propose a velocity-dependent compensation zone estimation method using Gaussian process that reduces interaction forces without compromising tracking performance. Our stability analysis shows that it can achieve uniformly ultimately bounded (UUB) behavior. The effectiveness of our approach is validated using a 6-DOF collaborative robot.