Annelids such as earthworms are considered to have central pattern generators (CPGs) that generate rhythms in neural circuits to coordinate the deformation of body segments for effective locomotion. At present, the states of earthworm‐like robot segments are often assigned holistically and artificially by mimicking the earthworms’ retrograde peristalsis wave, which is unable to adapt their gaits for variable environments and tasks. This motivates the authors to extend the bioinspired research from morphology to neurobiology by mimicking the CPG to build a versatile framework for spontaneous motion control. Here, the spatiotemporal dynamics is exploited from the coupled Hopf oscillators to not only unify the two existing gait generators for restoring temporal‐symmetric phase‐coordinated gaits and discrete gaits but also generate novel temporal‐asymmetric phase‐coordinated gaits. Theoretical and experimental tests consistently confirm that the introduction of temporal asymmetry improves the robot's locomotion performance. The CPG‐based controller also enables seamless online switching of locomotion gaits to avoid abrupt changes, sharp stops, and starts, thus improving the robot's adaptability in variable working scenarios.