Robotic communication with ants

Abstract

We used a robotic gantry to test the hypothesis that tandem running in the ant Temnothorax albipennis can be successful in the absence of trail laying by the leader. Pheromone glands were placed on a pin attached to a gantry. This set-up substituted for the leader of a tandem run. Neither the pin nor the glands touched the substrate and thus the ant following the robot was tracking a plume of airborne pheromones. The robot led individual workers from their current nest to a potential new one. The robotic gantry was programmed to allow for human intervention along its path to permit the following ant to stop and survey its surroundings and then catch up with its mechanical leader. The gantry then automatically tracked the precise route taken by each ant from the new nest back to the old one. Ants led by the robot were significantly more successful at finding their way home than those we carried to the new nest that had no opportunity to learn landmarks. The robot was programmed to take either a straight or a sinusoidal path to the new nest. However, we found no significant difference in the abilities of ants that had been led on such direct or sinuous paths to find their way home. Here, the robot laid no trail but our findings suggest that, under such circumstances, the following ant may lay a trail to substitute for the missing one.

Keywords: Animal–robot interaction; Learning; Orientation; Pheromones; Social behaviour; Tandem running.

Fig. 1.

The experimental arena was made of plexiglass with a matt white floor with dimensions 1134×980 mm (horizontal×vertical, not to scale). ON: old nest; NN: new nest; A: 50 mm mark; B: 350 mm mark (a further mark was made at 360 mm, i.e. 10 mm beyond the 350 mm mark, where the new nest was placed); Z: the zeroing spot (a pencil mark behind the new nest, on which the camera was centred to provide a consistent reference for the recorded movement coordinates). Blue and red lines: the path of the robotic leader in the straight and sinusoidal treatments, respectively.

Fig. 2.

The pin, C, was mounted at point A on an arm projecting from the gantry beam, and the camera was mounted at point B. This set-up was used both to lead the ants to the new nest (not shown) and to track them on their return journey to the old nest (also not shown).

Fig. 3.

Robot-led ants. (A) Survival function and (B) hazard function for the return duration (s) of ants led by the robot on a straight path, on a sinusoidal path or on a straight path with the arena cleaned before they embark on their return journey.

Fig. 4.

Ants carried manually to the new nest. (A) Survival function and (B) hazard function for the return duration (s) of ants manually placed at the new nest and tracked by the gantry on their return journey after the robotic leader was run (control) or not run (carried) on a straight or sinusoidal path without any followers.

Fig. 5.

Ants were occasionally observed clustered along the path taken by the gantry, perhaps responding to pheromone. The gantry had been run on (A) a sinusoidal path and (B) a straight path. Our results indicate that this surface pheromone does not originate from the robotic leader itself.

Fig. 6.

Analysis of path crossing and activity in each arena quadrant. The quadrants were labelled clockwise 1–4. The four examples given are: (A) a sinusoidal return: the ant was led by the robotic leader and allowed to return from the new nest; (B) a control carry: the ant was carried to the new nest but the robotic leader was run prior to its return without any followers; (C) a simulated run; and (D) a cleaned return: the ant was led by the robotic leader and the arena was cleaned of any pheromones before the ant began its return journey. ON: old nest (the area enclosed by the covering Petri dish; 120×120 mm); NN: new nest (the area of the vacant nest itself; 76×51 mm); area enclosed by the grey line around the filled grey rectangle of each nest: the nest for experimental purposes as this was the vicinity in which the ant could not normally be tracked; red line: the outward path of the robotic leader; green/blue line: the path taken by the returning ant in each quadrant; black dot: where a returning ant crossed the path of the outward robotic leader.

Fig. 7.

Number of crossings. The number of times the outward path of the robotic leader was crossed by ants returning to the old nest under different treatments. Red: robot-led ants; blue: carried ants; yellow: simulated ants (see Table 2). Box width is proportional to sample size.

Fig. 8.

Comparison between all robot-led ants and all carried ants. (A) Survival function and (B) hazard function for the return duration (s) of all robot-led ants (on a straight path, on a sinusoidal path and on a straight path with the arena cleaned before their return journey) and of all carried ants (manually placed at the new nest and tracked by the gantry on their return after the robotic leader was or was not run on a straight or sinusoidal path without any followers; see also Table 1).