Abstract
This document addresses computational redundancy in multi-agent systems where collaborating agents may independently generate isolated, natural language-based plans, potentially leading to the redundant execution of similar subtasks. The described systems and methods can provide a framework for multiple software agents to collaboratively plan and execute complex tasks by introducing a structured, joint planning phase. During this phase, agents may iteratively build a single task decomposition graph, proposing subtasks as nodes and defining prerequisites as directed edges. This unified, machine-readable plan can enable the system to identify and de-duplicate shared workloads. Furthermore, the explicit dependency structure can allow for the generation of an execution schedule, for example, through a topological sort, which may increase parallelism, reduce agent idle time, and improve computational efficiency.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Hartmann, Florian and Carbune, Victor, "Joint Task Graph Generation for Optimizing Multi-Agent Workloads", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/8824