Several anonymous Dubins-car like mobile robots travel in a planar environment that hosts a scalar field, like the level of radiation or concentration of a contaminant. The objective is to co-operatively detect and localize the boundary of the set where the field value exceeds a certain threshold. The robots suffer from deficits of competence, communication, perception, and maneuverability: they do not know the field profile a priory, are not aware of the team size, cannot communicate with and recognize one another, can measure only the value of the field at the current location, are subjected to nonholonomic constraints, and are able to move along paths of only limited curvatures. We propose a new decentralized navigation strategy that drives all robots to the desired environmental boundary, with subsequent stable circulation along it. This strategy is based on an autonomous control of every robot, prevents collisions between them and ultimately ensures their pseudo-uniform distribution over the boundary to better utilize the resources of the team for representatively portraying the boundary. Furthermore, the proposed control scheme does not employ gradient estimation, which typically needs ineffective concentration of robots into tight clusters, and is non-demanding with respect to both computation and motion. Its mathematically rigorous justification is provided. The effectiveness of the proposed guidance law is confirmed by computer simulations and real-world experiments.