Because animals are extremely effective at moving in their natural environments they represent an excellent model to implement robust robotic movement and navigation. Braitenberg vehicles are bio-inspired models of animal navigation widely used in robotics. Tuning the parameters of these vehicles to generate appropriate behaviour can be challenging and time consuming. In this paper we present a Reinforcement Learning methodology to learn the sensori-motor connection of Braitenberg vehicle 3a, a biological model of source seeking. We present simulations of different stimuli and reward functions to illustrate the feasibility of this approach.