Sviluppo di un modello software di sciami di robots per applicazioni di soccorso

In swarm robotics, the design of a group of robots often relies on distributed decision making without requiring any centralized control. This is confirmed also by nature, in which ant colonies, bee colonies, flocks of birds, schools of fish and human groups are best-known examples. These collective systems have proven to outperform single individuals, which are cognitive limited, in solving complex problems, such as resource allocation, foraging, defending from predators etc. This ability is the result of the collective decision making: individuals, while pursuing their individual goals, also adapt, being in a group, their behaviours to the action of their neighbours. In particular, social interactions promote information pooling among the agents, thus leading to the emergence of the collective intelligence (CI) of the group. Therefore, the CI is highly related to the collective behaviour and can’t be simply deduced by the aggregation of the individual cognitive levels. By modelling the dynamics of these multi-agent systems, the behaviour of a swarm of interacting robots, while solving a complex task, can be reproduced. Following this line of research, in this work, we employed the recently developed Decision Making Model (Carbone and Giannoccaro 2015; De Vincenzo et. al, 2017) to describe the dynamics of a swarm of individuals, engaged in rescue operations. They include rescue of victims in disaster zones, locate sources of hazards such as chemical or gas spills, toxic pollution, pipe leaks, radioactivity and so on. For these kinds of scenarios, which include protection or preservation of life, decentralized algorithms are often considered to be more robust than their centralized counterparts because there is no single point of failure, but this is rarely put to the test. Furthermore, differently from other algorithms and models (Dorigoet et. all., Kennedy et. all., Storn et. all., Karaboga et all, Krihnanand et. all., Yang X-S et. all., Mirjalili S et all.), the DMM is able to capture the main drivers of the decision making process: the self-interest and consensus-seeking. Therefore, the dynamics is controlled through two parameters: (i) the strength of social interaction βJ, (ii) the self-confidence of each agent β’. In this work, we employ a group of 7 individuals, searching for the highest human density zone on a square grid of 65536 points. Numerical simulations confirmed that exist some critical conditions, in terms of self-confidence and strength of social interactions values, leading to the emergence of the collective intelligence into the group. This state makes the group able to achieve its highest performance, i.e. the group fitness and degree of consensus among the agents, thus finding the highest human density zone. These results suggest, as further investigations, to make the Human Group Optimizer (De Vincenzo et. al., 2018) suitable for rescue operations.