Multi-robot task allocation has long been a fundamental challenge in large-scale autonomous systems due to the need for coordination, scalability, and robustness in complex environments. Traditional optimization-based or learning-based ap- proaches often struggle to maintain efficiency as task density and environmental complexity increase, while decentralized methods face communication bottlenecks and lack topological awareness. In this study, we introduce a topology-guided market-based allocation framework designed to improve coordination efficiency and task stability across multi-robot systems. Building on this representation, we formulate a market-based allocation protocol that leverages local bidding and con- sensus, augmented with topological stability metrics to balance travel distance, risk, and workload. We conduct systematic evaluations using the Birmingham dataset, focusing on the effects of feature persistence thresholds, anchor sampling ratios, and network size on allocation performance. The results identify optimal configurations that achieve consistent reductions in total travel distance, improved load balance, and enhanced allocation stability under uncertain conditions. This topology-guided market-based framework provides a scalable and interpretable foundation for large-scale, topology-aware coordination in multi-robot systems, bridging geometric representation learning with distributed decision-making.