The road construction industry is one which highly depends on just-in-time supply chains. The production planning and control on a building site is a complex process. In practice the planning is conducted well in advance so that the control function is limited by supervising the execution of the plan. The deviation between a predefined plan and its real-world execution is a common problem in the road construction industry. This master's thesis approaches this problem by developing a software component that executes the production planning as well as the production control ad-hoc. To build a dynamic and multifunctional system, a machine learning approach will be used for the implementation. This system will be analysed according to its usability in the real-world example of a road construction. As being a part of the cyber-physical system of the SmartSite research project, the resulting software component of this master's thesis will be executed on a real construction site.