Decades worth of experiments on the generation of virtual creatures have led to a large corpus of morphological and behavorial controllers. Yet despite these extensive researches, the role of the environment in the emergence of complexity is often experimenter-dependent. In this work, we present a framework in which environmental variables are the sole control mechanism for directing the evolution of an autonomous population of artificial creatures. The Environment-Driven Evolutionary Selection (EDEnS) algorithm automatically and simultaneously explores multiple alternatives, initially identical but with potentially differing dynamics with respect to their abiotic constraints. By exposing populations to different sets of constraints, they are subjected to divergent fitness functions both across the simulation space and between successive elementary steps. This framework is applied on a system composed of artificial plants autonomously reproducing in a 2D environment subjected to three factors: topography, temperature and hygrometry. We show that some of the populations obtained via EDEnS exhibit increased capabilities to invade foreign environments over populations evolved in a hospitable, constant environment. Moreover, the data thus collected highlighted two fundamental advantages of the automated exploration of abiotic constraints: the positive effects of catastrophical trimming and the unbiased designing of dynamical environments.