A hypothetical neuromusculoskeletal model is developed to simulate human normal walking and its modulated behaviors. A small set of neural periodic patterns drive spinal muscle synergies which in turn lead to specific pattern of muscle activation and supraspinal feedback systems maintain postural balance during walking. Then, the model demonstrates modulated behaviors by superimposing voluntary perturbations on the underlying walking pattern. Motions of kicking a ball and obstacle avoidance during walking are simulated as examples. The superposition of the new pulse command to a set of invariant pulses representing spino-locomotor is sufficient to achieve the coordinated behaviors. Also, forward bent walking motion is demonstrated by applying similar superposition. The composition of activations avoids a complicated computation of motor program for a specific task and presents a simple control scheme for different walking patterns.