The present work extends a two-level decision-making mechanism, modeling the cortico-basal ganglia (CBG) loop. It incorporates an exploration-exploitation control based on D1-type tonic dopamine (DA) effects in the corticostriatal synapses. The resulting model not only supports the previous findings reinforcing the feasibility of controlling the use of past information against exploring new options just by varying the level of D1-type tonic DA, but also shows how such control can increase lifetime while confronting a simple survival task. A MODI (MODular Intelligence) robotic platform is tested performing a standard survival task, proposing a robotics controller that integrates the CBG model as its action selection mechanism. The MODI robot has to deal with a two-resources survival problem, learning on-line which actions most likely offer reward for the agent, for a given time, in order to maximally extend lifetime. The obtained data shows relations between time survived and the level of D1-type tonic DA of the robot.