LN: Sha et al. (2026) — Multi-Agent RL Scheduling with State and Task Graph Structural Modeling

Bibliographic Reference

Citation: Sha, J., Song, M., Sui, G., Sun, H., & Dong, D. (2026). A multi-agent reinforcement learning scheduling algorithm integrating state graph and task graph structural modeling for ride-sharing dispatching. Scientific Reports, 16. https://doi.org/10.1038/s41598-026-35004-8

Important Note

Overview

The bibliography’s description (“applied to project scheduling”) is misleading. The verified paper is specifically about ride-sharing vehicle dispatch. The techniques (MARL + state graph + task graph) are transferable to PM scheduling, but the paper itself does not address project management.


Pass 1 — Bird’s Eye View (5 Cs)

CAssessment
CategoryAlgorithm proposal + empirical evaluation
ContextMARL applied to real-time vehicle dispatch optimisation
CorrectnessSimulation + real dataset. Strong experimental validation.
Contributions(1) State graph captures environment dynamics; (2) Task graph models dependency structure; (3) Combined graph neural network for MARL scheduling
ClarityTechnical. Clear algorithm descriptions.

Relevance: ⭐⭐⭐

The graph-based MARL approach is technically transferable to PM sprint scheduling (dependency graph of tasks). However, the paper is about ride-sharing, so this is a methodological transfer reference, not a direct PM paper.


PUMA Connection

Methodological reference for how to integrate task dependency graphs into a MARL scheduling agent. Relevant for PUMA Stage 5 if the Smart PMO implements a RL-based sprint planner. Reference for future work section (Stage 3 backlog prioritisation).

MOCs