Incentive for self-protection in a collective system: a swarm robotics case study

Designing controllers for agents of a collective system is challenging. The challenge lies in the nonlinearity between the behavior of the individuals and the emerging patterns in the collective. A branch of ongoing research in collective systems, e.g., in swarm robotics, concerns with discovering mechanisms that lead to specific collective behaviors indirectly, i.e., the effects of incentives in the emergence of interesting
collective patterns. Various intrinsic motivations have been suggested as the drivers of pattern development in natural and artificial collective systems. An example is the development of pathways to provide easier access to currents that flow through a system. Another example is the emergence of collective motion as a result of intrinsic motivation for maximization of potential future states. By getting inspiration from the principle of free energy minimization in biological systems, predictability of the future states have also been used as an intrinsic motivation, resulting in a number of collective formations. It demonstrated the tendency of agents to locate themselves in the positions that are less prone to changes in their surroundings. The current paper presents early investigations of the agents’ incentive for avoiding the surrounding environment by locating themselves between their peers, i.e., an incentive for self-protection. The swarm behavior resulted from this intrinsic motivation leads to formation of aggregates with high mobility of agents within them. The behavior
is loosely similar to the huddling behavior of emperor penguins, where the birds self-organize to take turn in locating themselves inside the crowd to stay protected from the wind. The behavior appears to show a relatively low sensitivity to the swarm density.