Reinforcement Learning Based Approach for Adaptive Release Planning in an Agile Environment

Agile programming involves continually evolving requirements along with a possible change in their business value and an uncertainty in their time of development. This leads to the difficulty in adapting the release plans according to the response of the environment at each iteration step. This paper shows how a machine learning approach can support the release planning process in an agile environment. The objective is to adapt the release plans according to the results of the previous iterations in the present environment . Reinforcement learning technique has been used to learn the release planning process in an environment of various constraints and multiple objectives. The technique has been applied to a case study to show the utility of the method. The simulation results show that the reinforcement technique can be easily integrated into the release planning process. The teams can learn from the previous iterations and incorporate the learning into the release plans.