Reinforcement Learning aided Optimal Resource Allocation Mechanism for Open Markets

Resource allocation is an extensively investigated problem in several market domains, not restricted to cloud computing platforms, supply chains, government procurement, etc. This dissertation primarily focuses on addressing challenges associated with resource allocation in open market settings. The open markets are characterised based on the behaviour of the participants, i.e., resource vendors and resource buyers. In such markets, participants have varying resource requirements, wherein they can enter or withdraw from the market dynamically. This leads to uncertainty in the resource availability and resource requirement in the market. But to ensure competitiveness and higher participation rate of the participants, the trade-off between and supply/demand in the market is crucial. Also, different participants have different sets of conflicting preferences. For instance, resource vendors aim to maximise their revenue, whereas resource buyers aim to minimise their costs. Thus, concurrently addressing such conflicting objectives increases the complexity of the resource allocation problem. Therefore, there is a need to design a resource allocation technique, popularly called a resource allocation mechanism (RAM), to address these challenges. This dissertation focuses on designing such RAMs for open markets in a preview of game theory. In this context, an efficient RAM depends on two basic rules, i.e., allocation rule and pricing rule. In doing so, this dissertation presents several RAMs adopting different pairs of custom-designed rules, called policies for resource allocation in open market settings.