Accurate and timely perception of collision in highly variable environments is still a challenging problem for artificial visual systems. As a source of inspiration, the lobula giant movement detectors (LGMDs) in locust's visual pathways have been studied intensively, and modelled as quick collision detectors against challenges from various scenarios including vehicles and robots. However, the state-of-the-art LGMD models have not achieved acceptable robustness to deal with more challenging scenarios like the various vehicle driving scenes, due to the lack of adaptive signal processing mechanisms. To address this problem, we propose an improved neuronal system model, called LGMD + , that is featured by novel modelling of spatiotemporal inhibition dynamics with biological plausibilities including 1) lateral inhibitions with global biases defined by a variant of Gaussian distribution, spatially, and 2) an adaptive feed-forward inhibition mediation pathway, temporally. Accordingly, the LGMD + performs more effectively to detect merely approaching objects threatening head-on collision risks by appropriately suppressing motion distractors caused by vibrations, near-miss or approaching stimuli with deviations from the centre view. Through evolutionary learning with a systematic dataset of various crash and non-collision driving scenarios, the LGMD + shows improved robustness outperforming the previous related methods. After evolution, its computational simplicity, flexibility and robustness have also been well demonstrated by real-time experiments of autonomous micro-mobile robots.

We present a new angular velocity estimation model for explaining the honeybee's flight behaviours of tunnel centring and terrain following, capable of reproducing observations of the large independence to the spatial frequency and contrast of the gratings in visually guide flights of honeybees. The model combines both temporal and texture information to decode the angular velocity well. The angular velocity estimation of the model is little affected by the spatial frequency and contrast in synthetic grating experiments. The model is also tested behaviourally in Unity with the tunnel centring and terrain following paradigms. Together with the proposed angular velocity based control algorithms, the virtual bee navigates well in a patterned tunnel and can keep a certain distance from undulating ground with gratings in a series of controlled trials. The results coincide with both neuron spike recordings and behavioural path recordings of honeybees, demonstrating that the model can explain how visual motion is detected in the bee brain. Author summaryBoth behavioural and electro-physiological experiments indicate that honeybees can estimate the angular velocity of image motion in their retinas to control their flights, while the neural mechanism behind has not been fully understood. In this paper, we present a new model based on previous experiments and models aiming to reproduce similar behaviours as real honeybees in tunnel centring and terrain following simulations. The model shows a large spatial frequency independence which outperforms the previous model, and our model generally reproduces the wanted behaviours in simulations. Introduction 1 Insects, like flies and honeybees, though with tiny brains can deal with very complex visual 2 flight tasks. It has been researched for decades how they detect visual motion. However, 3 the neural mechanisms behind for explaining varieties of behaviours including patterned 4 tunnel centring [1, 2] and terrain following [3-5] are still not very clear. According to the 5 honeybees' behavioural experiments performed, the key to their excellent flight control 6 ability is the angular velocity estimation and regulation [6,7]. For instance, honeybees fly 7