The methods mentioned here can provide efficient locomotion
for various types of modular robots. Several systems (e.g.
[28–30]) are well scalable for robots with many modules. However,
the performance of the locomotion may vary as the robots move
through environments with changing terrain. In these cases, the
parameters of the locomotion generator need to be adapted,
or another generator needs to be utilized. Moreover, a single
locomotion generator may not be suitable if obstacles need to
be avoided. An approach utilizing several gaits that are switched
according to the type of terrain was proposed in [34]. This kind of
model contains many parameters that need to be tuned in advance.
This is not suitable for modular robots, which can reconfigure
to many shapes. Another approach for adapting the locomotion
utilizes sensory information in the feedback [25,35–37]. However,
like the previous approach, it is difficult to optimize the required
parameters.