Bugs with Features: Resilient Collective Motion Inspired by Nature

In collective motion, individuals move in an ordered manner, without centralized control, limited to highly localized perception. While natural collective motion is robust, most artificial swarms are brittle, and can be disrupted when robots perceive others unreliably (as is common in realistic perception , e.g., visual). This paper presents mechanisms for robust collective motion, inspired by studies of locust behavior. First, we circumvent the challenge of estimating the (lack of) motion of others in realistic settings, by introducing intermittent pauses in robot motion. During pauses, robots can reliably perceive neighbors’ velocities, and isolate failing neighbours from affecting decisions once motion resumes. However, a robot pausing to perceive also risks being perceived by others as failing. We therefore explore the complex relation between the duration of pauses, and the isolation behavior. Second, we show that erroneous distance estimations disrupts collective motion. We show that combining visually perceived horizontal and vertical sizes of neighbors corrects these estimations, even under occlusions. Through extensive physics-based simulations experiments, we show dramatic improvements to swarm resilience when using these techniques.