The paper introduces Lag, a policy designed for online task assignment with reusable resources. In this setting, tasks arrive over time and must be matched to resources that become available again after finishing a task. The authors frame this as a general problem with practical reach, citing peer reviewer assignment, skill-based call routing, rideshare matching, emergency dispatch, and crowdsourcing as examples.
The abstract specifies that tasks arrive according to a geometric process, though further details on the arrival model are cut off in the available text. The central claim is that Lag achieves mean-field optimality, meaning it performs optimally in the limit of large systems, a common criterion for such stochastic control problems.
Because only the abstract is available, the article does not describe the policy's implementation or numerical results. The contribution, as stated, is a provably optimal policy for a broad class of reusable-resource assignment problems.