In this paper, formation adaptation and stability for mobile multi-agent systems is studied. The objective is to suggest a set of robust control functions that can be combined to build complex formations of mobile robots. A properly designed formation should be able to follow a single leader in a clustered environment and to, if necessary, adapt to the surroundings by changing its shape. The two control algorithms proposed here are adapted for systems with limited communication capacity and low performance sensors. The algorithms only require information that can be directly achieved from on-board sensors and, in particular, they only need very coarse estimation of the velocity of the neighbors in the formation. An arbitrary change of the shape of a formation may require switching between control algorithms. In the paper, it is shown that switching between the two proposed control algorithms is stable under some reasonable assumptions. The results are verified by simulations, which also show that switching can be performed safely even with a high noise level and no prior filtering of sensor input.