A simple strategy for detecting moving objects during locomotion revealed by animal-robot interactions

Abstract

An important role of visual systems is to detect nearby predators, prey, and potential mates, which may be distinguished in part by their motion. When an animal is at rest, an object moving in any direction may easily be detected by motion-sensitive visual circuits. During locomotion, however, this strategy is compromised because the observer must detect a moving object within the pattern of optic flow created by its own motion through the stationary background. However, objects that move creating back-to-front (regressive) motion may be unambiguously distinguished from stationary objects because forward locomotion creates only front-to-back (progressive) optic flow. Thus, moving animals should exhibit an enhanced sensitivity to regressively moving objects. We explicitly tested this hypothesis by constructing a simple fly-sized robot that was programmed to interact with a real fly. Our measurements indicate that whereas walking female flies freeze in response to a regressively moving object, they ignore a progressively moving one. Regressive motion salience also explains observations of behaviors exhibited by pairs of walking flies. Because the assumptions underlying the regressive motion salience hypothesis are general, we suspect that the behavior we have observed in Drosophila may be widespread among eyed, motile organisms.

Figure 1

Cartoon illustrating the principle of regressive motion salience. (A) The stationary animal at the center has no difficulty using optic flow to distinguish moving objects (flies) from stationary objects (circles) in the background. Thick green arrows indicate motion of flies and thin red and black arrows indicate the angular velocity of edges subtended by objects on the retina of the fly at center. (B) An animal moving in a straight line without rotation will experience progressive optic flow of all stationary objects in its environment. The magnitude of angular optic flow (indicated by black arrows) that each object creates will depend on its distance to the fly and its orientation relative to the direction of motion. An object that moves in such a way as to create progressive optic flow (i) will be difficult to distinguish from the apparent motion of the stationary background, whereas an object moving so as to create regressive optic flow (ii) will be much easier to detect.

Figure 2

Encounters between walking fruit flies provide support for regressive motion salience. (A) Cartoon illustrating the definitions for ϕ, the azimuthal position of one fly in another’s reference frame and θ, the angle subtended by the other fly. (B–E) Four example encounters between two flies. Fly-shaped icons in the top panels illustrate the position of the two flies at 100 ms intervals. Initial direction is indicated by small arrows and a white fill color indicates the position of both flies when the red fly stops. The middle panels plot the translational speed of the two flies. The lower panel plots the time course of ϕ and θ. In these plots, time runs from top to bottom on the vertical axis and the θ is indicated by the width of each point as plotted along the abscissa. Red indicates images experienced by the red fly, blue indicates images experienced by the blue fly. Just prior to stopping, the red flies experience regressive motion, as indicated by the fact that the slope of azimuthal position slants toward the midline (black arrow) (B,C). Examples of encounters in which two flies walk on a collision course (‘T-stops’). The fly that would arrive at the intersection point first perceives progressive motion and continues without stopping. The slower fly, which stops, perceives regressive motion. (D,E) Examples of encounters in which two flies walk along parallel courses (‘drag races’). In these cases, the faster of the two flies perceives progressive motion while the slower fly, which stops, perceives regressive motion. The examples were selected from a previously published data set[7].

Figure 3

Flyatar: a simple fly-sized robot that may be programmed to interact with a real fly. (A,B) CAD drawings of the apparatus illustrating the motors and drives that control the position of a sub-stage magnet that actuates the small robot above the stage. In A, the arena and circular thermal barrier have been removed to show the sub-stage motor system. (C) Photograph of the apparatus showing the checkerboard background around the arena and the lighting system. The mirror array helped provide even lighting conditions. (D) Cartoon showing size and shape of the robot compared to a real fly.

Figure 4

Evidence for regressive saliency in fly-robot interactions. (A–D) Example trajectories in which the robot was programmed to start moving when a fly walked past. Robot (red circle) and fly (blue fly-shaped icons) indicate positions at 200 ms intervals. The frame in which the robot started moving is indicated in black fill and by a solid line. The traces below each sequence indicate the speed of the fly and robot throughout the encounter. (A) Example in which the robot creates progressive motion. (B) Another example of progressive motion. Note that in this case, the robot moves forward in the direction of fly motion, but the speed differential is such that the image of the robot was still progressive in the fly’s reference frame. (C,D) Examples in which the robot creates regressive motion. Note that the flies stop walking (indicated by white fill and dotted line) almost immediately following the onset of robot motion. (E) Histogram of all valid encounters (n=10,047) from 46 flies. Values along the ordinate axis are plotted on a log scale to better visualize rare events. Gray area indicates distribution of stimulus angular velocity for all trials and the superimposed black area indicates the distribution of trials in which the flies stopped within 850 ms after the onset of robot motion. (F) Stop probability (calculated as the ratio of the distributions plotted in panel E) as a function of angular velocity when the robot was present (closed circles) and for the ‘no robot’ controls. P_stop values for angular velocities above +120 °/sec and below −100 °/sec are not displayed because the small sample size in these ranges render the calculated ratios unreliable. In ‘no robot’ controls (n=3528 trials, 14 animals), the robot was removed from the arena but the sub-stage actuators and control software operated as in normal trials. (G) Similar plot to that in panel F, but in this case the traces represent a population average and SEM envelopes for the probability functions evaluated for each individual fly (N=46). (H) The data in F are replotted after parsing trials in five groups according to the absolute angular velocity of the fly at the start of the trial: (0<|ω|<10, 10<|ω|<20, 20<|ω|<30, 30<|ω|<40, 40<|ω|). Each group is plotted as a different tone from red to black. (I) Stop probability as a function of stimulus angular velocity for data parsed according to the initial position of the robot at the start of motion. The probability curves are similar for motion initiated in the front and rear sectors of the visual field. (J) Stop probability plotted as a function of the distance between the fly and robot at the start of robot motion for progressive (open) and regressive (closed) motion. The curves are derived from the range of regressive angular velocities that evoked the peak response (−100 to −60 °/sec), and each point averages trials across a range of angular velocities.