Cockroaches orient towards and track plumes of airborne odours using simple yet sophisticated processing of odour concentration stimuli. Their orientation behaviour is driven not only by instantaneous sampling of odour concentration but by temporal dynamics of the odour signal which are processed by specialised On/Off olfactory receptor neurons encoding rate of change of odour concentration. We present a neurorobotic model for chemotaxis that implements such sensory processing. The model consists of a sensorimotor connection embodied in a simulated Braitenberg vehicle with a dynamic sensory processing variant which extends existing functional models, and the model parameters are tuned via an optimisation algorithm. Simulation results demonstrate that the vehicle with the contribution of temporal dynamics and optimised parameters is able to localise the odour source in both a simplified odour plume characterised by a steady state Gaussian distribution of odour concentration and a realistic odour plume based on filament-based gas dispersion modelling.