Completing a project on time and within budget is not an easy task, and both academia and business have spent a lot of attention on this challenge. In order to cope with the uncertainty, project managers include time and cost buffers in the project to absorb as much unexpected events as possible. However, often a single centralised project buffer is used, which is highly inefficient and rigid. Therefore, our research will rely on multiple decentralised activity buffers that are constructed using a large amount of information derived from activity attributes. This research proposal consists of four studies. The first two studies aim at clustering the activities to optimally assign a limited number of buffers to groups of activities as well as to optimise the allocation of buffers to the project schedule. The next two studies optimally use the buffers for monitoring the project performance and improving the corrective actions under a wide set of management restrictions.