This paper discusses experimental verifications and numerical analysis of a two-dimensional modular robot called "Slimebot", consisting of many identical modules. The Slimebot exhibits adaptive reconfiguration by exploiting a fully decentralized algorithm able to control its morphology according to the environment encountered. One of the significant features of our approach is that we explicitly exploit "emergent phenomena" stemming from the interplay between control and mechanical systems in order to control the morphology in real time. To this end, we particularly focus on a "functional material" and a "mutual entrainment" among nonlinear oscillators, the former of which is used as a spontaneous connectivity control mechanism between the modules, and the latter of which acts as the core of the control mechanism for the generation of locomotion. Experimental results indicate that the proposed algorithm can induce locomotion, which allows us to successfully control the morphology of the modular robot in real time according to the situation without losing the coherence of the entire system.