A Hybrid MILP–RL Framework
for Scaled Agile Projects
PEER REVIEWED PAPER
By Enoch Oghene-Mairo OMAJEH
Port Harcourt, Nigeria
Abstract
Sprint planning in large-scale agile projects is a complex decision-making process that involves balancing task priorities, team capacity, skill alignment, and evolving project conditions. Traditional approaches rely heavily on expert judgment or static optimization models, which often struggle to adapt to dynamic environments. This paper proposes a hybrid sprint planning framework that integrates Mixed-Integer Linear Programming (MILP) with reinforcement learning (RL) to generate feasible and adaptive task allocation strategies. The MILP model ensures optimal task allocation under capacity, skill, and dependency constraints, while the RL component learns from execution feedback to refine decisions across successive sprints in a closed-loop manner. The proposed approach was evaluated using a dataset comprising approximately 500 tasks and 12 team members, with historical sprint data used as a real-world baseline. Results show that the framework consistently outperforms traditional project management outcomes, achieving improvements in task allocation quality (+6.17%), workload balance (−27.78% variance), and delivery reliability (−57.14% task spillover), while maintaining near-real-time optimization performance (≈1.94 seconds). Furthermore, simulation results demonstrate that the RL component progressively improves allocation quality over successive sprints, increasing the normalized objective score from 0.86 to 0.93. These findings demonstrate that integrating optimization with adaptive learning enables not only high-quality initial sprint plans but also continuous performance improvement over time, making the approach well-suited for dynamic agile environments.
Keywords: Adaptive Sprint Planning; Mixed-Integer Linear Programming; Reinforcement Learning; MILP-RL Framework; Hybrid Optimization; Agile Project Management; Task Allocation
1.0 Introduction
1.1 Agile Methodologies
Agile methodologies have become the go-to approaches for managing complex and rapidly evolving projects. This is because they enable teams to respond effectively to changing requirements by breaking down project development into several short development cycles that allows continuous feedback (Highsmith, 2009; Rigby et al., 2023).
However, when organizations adopt agile practices across large and geographically distributed teams, the challenge of coordinating work at scale become pronounced. Scaled agile frameworks, such as the Scaled Agile Framework (SAFe) and Large-Scale Scrum (LeSS) were developed to provide structural guidance for multi-team coordination, yet they continue to rely heavily on manual planning and human judgment.
1.2 Sprint Planning in Large-scale and Distributed Environments
Sprint planning is a critical activity in agile project management, during which tasks are selected, prioritized, and assigned to team members based on specific criteria (which include capacity, skills, and dependencies).
As the number of teams grows, this process becomes more and more complex due to heterogeneous team capabilities, inter-task dependencies, fluctuating workloads, and evolving project priorities (Larman & Vodde, 2016).
Traditional planning approaches, which are based on expert judgment and simple heuristics, are often insufficient to handle this level of complexity. And organizations using them frequently experience suboptimal workload distribution, skill mismatches, and resource bottlenecks (Larman & Vodde, 2016).
More…
To read entire paper, click here
How to cite this paper: Omajeh, E, O-M. (2026). Adaptive Sprint Planning: A Hybrid MILP–RL Framework for Scaled Agile Projects; PM World Journal, Vol. XV, Issue V, May. Available online at https://pmworldjournal.com/wp-content/uploads/2026/05/pmwj164-May2026-Omajeh-Adaptive-Sprint-Planning-in-Scaled-Agile.pdf
About the Author
Enoch Oghene-Mairo Omajeh
Port Harcourt, Nigeria
Enoch Oghene-Mairo OMAJEH is a Project Planner and a PhD candidate in Information Systems Engineering in the University of Port Harcourt, Nigeria (expected in 2027). With vast experience leading project execution and digital transformation, his interest lies in the intersection of Project Management and IT, particularly Artificial Intelligence. His focus is to contribute to the development of intelligent systems that support project decision-making, enhance execution efficiency, and improve overall project delivery performance. He can be contacted at eomajeh@gmail.com
