Restoring motor function is the primary goal of neurorehabilitation, but direct gait training usually involves risks. As a safe alternative, the effectiveness of pedaling exercise depends on its similarity to walking in terms of neuromuscular control strategies. This study aimed to quantitatively compare the neuromuscular control patterns of walking and pedaling under different conditions using muscle synergy analysis, with a focus on exploring the effect of dynamic loading on the effectiveness of simulated walking. This study utilized eight lower-limb muscles and 3D motion data from healthy participants under three conditions (walking, constant-load pedaling, and sinusoidal-load pedaling) to extract four muscle-synergy patterns using non-negative matrix factorization (NNMF). After matching the similarity of W vectors, we compared the sequence and waveform similarities of their activation patterns (H-vectors). The results revealed three shared synergistic patterns and one task-specific synergistic pattern between walking and the two pedaling conditions. This study demonstrated that pedaling movements could reproduce the basic neuromuscular modules of walking. More importantly, the findings suggest that dynamic loading is an effective strategy for faithfully simulating the temporal control patterns (timing and sequence) of walking, providing key insights for the control design of lower-limb rehabilitation robots to maximize therapeutic effects.