It is argued that traditional models of urban development are characterised by an aggregate mechanistic description of statistical units. Furthermore, important aspects of transportation are not included in these models, but urban development can be regarded as a combined process of land use change, transportation system and lifestyles. New developments in evolutionary theory provide a new paradigm for a microsimulation approach on the level of individuals, which accounts for diversity, learning and change in the population of the modelled system. In this thesis a framework for agent-based simulations will be presented for which this new evolutionary theory provides the theoretical background. The essence of the approach builds on the mutual interdependencies between all system elements, in this case inhabitants and their environment. This principle is extended to change in the interactions of the system over time, leading to an adaptive system that mutually specifies all its elements over time. On this framework an adaptive agent-based model for the use in urban simulations is built. The agents are equipped with a set of intrinsic needs, the satisfaction of which is expressed through a set of corresponding budgets. The budget state is fed into a Fuzzy Logic rule base for decision making. As opposed to many existing approaches to microsimulation, the agents are designed to change their behavioural rules during run time according to experience. Different adaptation strategies are tested and compete with each other. The results of the model vindicate the conceptual framework. The essence of the underlying theory - mutual specification based on satisficing as opposed to optimisation - leads to a cognitive approach to the simulation of socio-natural systems. Microsimulation based on adaptive agents can help integrate many aspects of urban models, which are conventionally treated by separate models and can help clarify the implications of change for the inhabitants of an urban system.