Predicting the Behavior of Robotic Swarms in Search and Tag Tasks

The Blessing and Curse of Emergence in Swarm Intelligence Systems

“…In the context of swarm robotics examples include: Lancaster who uses networks of simple probabilistic graphs to predict swarm behaviour [38]; Dixon et al who have investigated the verification of swarms using temporal logic and model checking [39]; and Brambilla et al who have introduced an approach to the top-down design and verification of swarms via formal specification and model checking [40]. Less progress has been achieved on validation of safety aspects but Harper has shown the potential for using Lyapunov stability techniques [41].…”

A Data-Driven Soft Sensor for Swarm Motion Speed Prediction Using Ensemble Learning Methods

“…cise controllers for individual robots and understanding the swarms' full behavior leading to their design [8].…”

“…Generally speaking, predictive models at the microscopic level focus practically on capturing the dynamics of swarm components and their interactions and their environment [2], [8]. Although with these models, precise behavior predictions are achievable.…”

“…However, due to the apparent tracking of individuals and their interactions, they are not sufficiently scalable when the size of the swarm is sizeable [9]. On the other hand, predictive models at the macroscopic level consider the swarm as a whole and thus provide a high-level abstraction of the swarm dynamics, resulting in their high adaptability to swarms with large sizes [8], [10].…”

“…Many of the macroscopic predictive models of collective swarm behaviors are part of a wide range of mathematical models. Most of them use sets of differential equations to track swarm properties associated with the swarm collective behaviour [11], [12]. Despite the valuable insights that these models have brought to understand swarm behaviors, however, these models suffer from the likelihood of losing the aptitude to seizure outliers hidden in the process [11], [13].…”