We present an evolutionary robotics investigation into the metabolism constrained homeostatic dynamics of a simulated robot. Unlike existing research that has focused on either energy or motivation autonomy the robot described here is considered in terms of energy-motivation autonomy. This stipulation is made according to a requirement of autonomous systems to spatiotemporally integrate environmental and physiological sensed information. In our experiment, the latter is generated by a simulated artificial metabolism (a microbial fuel cell batch) and its integration with the former is determined by an E-GasNet-active vision interface. The investigation centres on robot performance in a three-dimensional simulator on a stereotyped two-resource problem. Motivationlike states emerge according to periodic dynamics identifiable for two viable sensorimotor strategies. Robot adaptivity is found to be sensitive to experimenter-manipulated deviations from evolved metabolic constraints. Deviations detrimentally affect the viability of cognitive (anticipatory) capacities even where constraints are significantly lessened. These results support the hypothesis that grounding motivationally autonomous robots is critical to adaptivity and cognition.