In this paper we discuss the experimental implementation of a chemical reaction-diffusion processor for robot motion planning in terms of finding the shortest collision-free path for a robot moving in an arena with obstacles. These reaction-diffusion chemical processors for robot navigation are not designed to compete with existing silicon-based controllers. These controllers are intended for the incorporation into future generations of soft-bodied robots built of electro- and chemo-active polymers. In this paper we consider the notion of processing as being implicit in the physical medium constituting the body of a 'soft' robot. This work therefore represents some early steps in the employment of excitable media controllers. An image of the arena in which the robot is to navigate is mapped onto a thin-layer chemical medium using a method that allows obstacles to be represented as local changes in the reactant concentrations. Disturbances created by the 'objects' generate diffusive and phase wave fronts. The spreading waves approximate to a repulsive field generated by the obstacles. This repulsive field is then inputted into a discrete model of an excitable reaction-diffusion medium, which computes a tree of shortest paths leading to a selected destination point. Two types of chemical processors are discussed: a disposable palladium processor, which executes arena mapping from a configuration of obstacles, given before an experiment and, a reusable Belousov-Zhabotinsky processor which allows for online path planning and adaptation for dynamically changing configurations of obstacles.