On the Performance of Altruistic Infotaxis Under Erroneous Communication Channel

Locating sources of hazardous phenomena (or undesired behaviour such as satellite jamming) accurately and timely is key to emergency response operations (or operation under hostile environment). Classically, odor-based localization is performed using trained animals with a sharp sense of smell (e.g. dogs). Instead, the localization of a toxic gas leakage source (or nuclear radiation source) can be performed by robot searchers (agents) equipped with proper sensors to avoid endangering animals. Typically, the leakage distribution encounters severe randomness and sparsity induced by the diffusion physical model and the environmental aspects (e.g. varying flow currents in the medium). As a consequence, classical gradient-based search algorithms become inefficient. Infotaxis has been proposed as an effective non-gradient-based method for source localization. In this paper, we study the case of two altruistic search agents leveraging Infotaxis assuming non-ideal (erroneous) conditions for the communication channel used for cooperation. The performance of the altruistic agents is examined from accuracy and efficiency points of view. The results show that although having multiple agents improves the average search time, a severely degraded communication channel causes the rate of successfully finding the source to drop drastically, suggesting a non-cooperative approach in such scenarios.