Towards Collective Visual Perception in a Multi-agent Model of Slime Mould

Sensation and perception of the surrounding environment is an essential mechanism in enhancing survival by increasing opportunities for foraging, reproduction and avoiding predatory threats. The most complex and well developed sensory mechanism is vision, which is highly developed in mammalian neural systems. Simple organisms, such as the single-celled slime mould Physarum polycephalum possess no neural tissue yet, despite this, are known to exhibit complex computational behaviour. Could simple organisms such as slime mould approximate complex perceptual phenomena without recourse to neural tissue? We describe a multi-agent model of slime mould where complex responses to the environment such as Lateral Inhibition (LI) can emerge without any explicit inhibitory wiring, using only bulk transport effects. We reproduce the characteristic edge contrast amplification effects of LI using excitation via attractant based stimuli. We also demonstrate its counterpart behaviour, Lateral Activation (where stimulated regions are inhibited and lateral regions are excited), using simulated exposure to light irradiation. Long-term changes in population density distribution correspond to a collective representation of the global brightness of 2D image stimuli, including the scalloped intensity profile of the Chevreul staircase and the perceived difference of two identically bright patches in the Simultaneous Brightness Contrast (SBC) effect. We demonstrate a realistic perception of a greyscale scene generated by the movement trails of the agent population and explore how an artistic sketch-like perception can be achieved by purposefully distorting the sensory inputs to the agent population. This simple model approximates Lateral Inhibition, global brightness perception, and thus primitive vision, in a collective unorganised system without fixed neural architectures. This suggests novel collective mechanisms and sensors for use in distributed computing and robotics applications.