Search for a command to run...
The SonoRo is a swarm robotics platfrom that can emit and receive sound to perform collective behaviour. By leveraging off-the-shelf hardware and readily available software, SonoRo aims to provide an onboard, scalable and easily reproducible platform tailored for animal and robot swarm studies. Here the data from four different tasks is presented: collision avoidance with static sound sources, goal search a static environment, multirobot aggregation, and multirobot dispersion. Each experiments contains raw video data, robot trajectories, video with robot position tracked and summary results. Experiments' description We conducted four types of experiments to test SonoRo's ability to perform simple tasks and to validate the system in collective settings. In single robot experiments, we tested goal search and source rejection, aimed at demonstrating individual performance in sound localisation. In experiments with three SonoRo robots, we tested aggregation and dispersion to demonstrate our platform's potential for swarm robotics. In all scenarios, we ran the experiments in a rectangular arena of 2×1.4 m delimited by white reflective tape. The arena was located approximately in the middle of a room sized 7×4.5 m. The experiment summary is shown in the SonoRo_ANTS_final.mp4 Individual Experiments Goal search The task consisted of locating the loudspeaker by tracking the emitted chirps. In this experiment, there was one active loudspeaker raised 10 cm off the ground, located near one corner of the arena. The loudspeaker emitted 20-ms chirps sweeping from 9 to 4 kHz. The SonoRo's own speaker was silent, and the robot was only acquiring audio data. The sound-intensity threshold to initiate an attractive response on the SonoRo was set to 65 dB SPL RMS (ref. 20μPa). The SonoRo responded to the detected sound by turning by a 5-degree angle towards the estimated peak-angle direction, or by moving forward when the estimated peak-angle direction was located in front of SonoRo in a range of ±10 degrees. We performed eight independent trials, each beginning with the SonoRo positioned at a random location and orientation within the arena, at least 80 cm from the loudspeaker. Once the robot was close enough to the loudspeaker (15 to 20 cm), the trial was manually stopped, leading to trials with durations ranging from 25 to 60 seconds. Video files: 2025-11-15_19-13-252025-11-15_19-15-552025-11-15_19-23-052025-11-15_19-25-262025-11-15_19-27-142025-11-15_19-29-022025-11-15_19-31-362025-11-15_19-49-41 Avoidance The avoidance task was designed to evaluate the obstacle-avoidance capabilities of SonoRo. The SonoRo's own speaker was silent, and the robot was only acquiring audio data. In this experiment, five active loudspeakers, raised 10 cm off the ground, were positioned at different locations in the arena . White reflective tape was placed around each loudspeaker, creating a 10-cm radius buffer zone, and all cables were also taped to the arena floor to prevent entanglement with SonoRo's wheels. All loudspeakers simultaneously played 20-ms chirps sweeping from 9 to 4 kHz with equal intensity. The objective of SonoRo was to avoid colliding with the loudspeakers. The sound-intensity threshold that triggered a dispersive response was set to 70 dB SPL RMS. When this threshold was exceeded, SonoRo executed the DAS algorithm and turned 5 degrees away from the estimated peak-angle direction. We performed three independent trials, each lasting between 5 and 20 minutes, allowing the SonoRo to explore the arena extensively. Video files: 2025-11-14_19-02-582025-11-14_19-35-462025-11-14_19-44-44 Collective Robotics Experiments We used three SonoRo robots running identical controllers. We tested two different scenarios where the robots were programmed differently to perform either the aggregation or dispersion task. Each robot signalled its location to others by emitting chirps repeatedly. These two tasks tested the SonoRos' fundamental skill of localising each other. In both tasks, the three SonoRos moved in an empty arena. In each trial, the robots were placed in the arena with random orientations and positions. To avoid trivially easy initial conditions, the starting configuration was chosen to be sufficiently distant from the target configuration. Aggregation This experiment tested the ability of the SonoRo units to aggregate inside the arena. Here, the robots were positioned so they were neither facing each other nor facing the centre of the arena. Each SonoRo unit had an attraction and repulsion sound-intensity threshold. The attraction threshold allowed the robots to find each other and aggregate, while the repulsion threshold enabled them to avoid colliding when in close proximity. For all trials the SonoRo units were attracted to sound levels ≥ 65 dB SPL RMS. The tested repulsion thresholds and steering angles varied across trials. In three trials, the robots were repelled at sound levels ≥ 85 dB SPL RMS, and used a steering angle of 10 degrees. In the remaining six trials, the repulsion threshold was set to ≥ 80 dB SPL RMS and the steering angle to 5 degrees. Despite the difference in parameters used, the qualitative behaviour of the SonoRo groups remained the same, and thus we chose to present them together. In total, nine independent trials were performed, with durations ranging from 70 to 110 seconds. Video files: 2025-11-17_14-01-302025-11-17_14-06-032025-11-17_14-08-462025-11-17_14-35-252025-11-17_14-39-122025-11-17_14-44-312025-11-17_14-50-592025-11-17_14-54-072025-11-17_14-57-47 Dispersion This task consisted of the robots moving away from one another to spread out within the arena. This was the most challenging task because the SonoRo hardware design exhibits a frontal bias in both its microphones and loudspeaker. To improve localisation when robots faced in opposite directions, we increased the speaker output level by 4 dB compared to the aggregation experiment. The higher speaker level ensured mutual chirp detection on all members. The repulsion sound level was set to ≥ 60 dB SPL RMS. The steering angle of the SonoRo units was five degrees. The three robots were placed in close proximity, ensuring they were not facing in opposite directions. This arrangement forced the robots to first turn, and perform collision avoidance before dispersing into the wider arena. Six independent trials were performed with durations lasting between 110 to 180 seconds. Video files: 2025-11-17_15-37-132025-11-17_15-41-252025-11-17_16-07-342025-11-17_16-10-322025-11-17_16-13-072025-11-17_16-16-36 Results Individual Results single_attraction_subplot.png shows the results of eight trials of the goal search experiment. The plot shows how the distance from the target changes over time. In all trials, SonoRo successfully reached the goal while maintaining a constant speed, as evidenced by the linear slope. In the initial phase of each trial, the robot's distance from the goal often increases or remains constant for a few seconds. This occurs because the robot was initialised facing away from the goal, and it may require some exploratory movements away and towards the source before reaching the attraction sound-intensity threshold and triggering a behavioural response. By analysing the trajectories, we observed that the robot tends to move in a straight line towards the goal, maintaining a steady heading once the goal is localised. single_repulsion_subplot.png shows the results from the experiments in the avoidance tasks. The SonoRo successfully avoided multiple acoustic obstacles while moving randomly in the environment for several minutes. By observing the trajectory from a representative trial, depicted in the left panel (cyan line), we can infer that the SonoRo, in addition to modifying its trajectory when crossing the protective tape around the arena, also actively avoided the loudspeakers. The right panel shows the distance from the closest speaker throughout each of the three trials. The mean distance ( black dot in the plots ) for all trials was approximately 0.4 m (10–90th percentile range 0.27–0.53 m shown as whiskers). Despite the sequential nature of DoA computation, the 5 Hz sensory update rate allows the system to resolve overlapping signals robustly. Collective Results Across all nine trials of the aggregation experiment, the three identical SonoRo robots successfully aggregated in a common location, acting as mutual acoustic beacons. As illustrated in the right panel of multi_attraction_subplot.png, the robots consistently reduced the distance from one another following a brief search phase. We quantified the performance of such an aggregation behaviour by computing the convex polygon perimeter of the robot centres. The perimeter rapidly drops after a short initial plateau or increase, confirming a robust aggregation behaviour independent of the starting configuration. The polygon perimeter stabilizes towards the end of the trials, indicating a steady state. The final perimeter value of approximately 70 cm is a result of the hard-coded sound-intensity repulsion threshold. The SonoRo robots performed the dispersion task successfully in all trials. To ensure active avoidance rather than passive divergence, the robots were initialised in close proximity and facing one another. multi_dispersion_subplot.png confirms the repulsion behaviour: the representative trajectories in the left panel exhibit turning manoeuvres rather than straight paths, while maintaining a general outward expansion. In the right panel, the convex polygon perimeter metric shows a rapid initial increase. Notably, there is no initial decrease in the metric, despite the inward-facing start positions, confirming the robots' rapid response in being repelled by one another (aided by the front-facing microphones). Similar to the aggregation case, the metric stabilizes, this time at around five metres. The five-metre plateau is determined