The evaluation of pollutant levels is a key aspect on the issue of keeping a clean environment. Conventional techniques include the utilisation of a fixed setup incorporating pollutant sensors. However, these approaches are a very long way from an accurate monitoring. Thus, to improve pollutant monitoring on a power plant chimney, the use of robotic agent societies (mobile robots) is suggested. This suggestion is adequate in pollutant monitoring when the environment is hostile and/or the region to be sampled has large dimensions. However, the implementation of a system incorporating robotic agents raises complex technological problems. Before a set of any kind of real robotic agents is implemented, an accurate evaluation must be performed. What this paper describes is a simulated application of small flying robotic agent societies (helicopter models) monitoring a pollutant cloud. This simulation intends to show that an “intelligent” search method works better than a systematic or random procedure. In this kind of environment (dynamic and non-structured) and using mobile robotics to meet a goal such as this, a behavioural control architecture seems to meet the performance objectives. The behaviours designed to control the agents are prepared to implement individual needs (survival and navigation) and social needs (follow or gather group). The agents as individuals are capable of performing such a mission, however, global results are enhanced by social strategies.