This work deals with theoretical aspects of creating purposeful self-organization in collective autonomous systems. The core of this approach is closely related to a qualitative dynamics of corresponding models and to the reductive synergetic approach. Changes of qualitative dynamics can be associated with, for instance, collective decision making. Through controlling qualitative dynamics by using different control approaches, it is possible to create dedicated decision-making mechanisms for real autonomous systems. An essential aspect of the work is devoted to the design of interactions. It is demonstrated that spatial interactions, during collective decision-making, can be represented as coupled map lattices with fixed or random couplings. By using the normal form reduction, or the center manifold approach, lattices can be reduced to "order parameters." This work considers three different forms of order parameters, which correspond to decision-making processes such as voting, bargaining, and selection. These decision processes are implemented in two real collective systems: first, agent-based process planning in an industrial environment, and second, in a swarm of micro-robots during collective energy foraging. The work demonstrates the advantages of applying analytical mechanisms to micro-systems: simplicity and reliability of decision mechanisms, minimal requirements on hardware resources and energy consumption, and the predictability of collective dynamics.