Autonomous materials research systems allow scientists to fail smarter, learn faster, and spend less resources in their studies. As these systems grow in number, capability, and complexity, a new challenge arises—how will they work together across large facilities? We explore one solution—a multi-agent laboratory-control framework. The framework is demonstrated with autonomous materials science labs in mind, where information from diverse research campaigns can be combined to address scientific questions. The framework can (1) account for realistic resource limits, e.g., equipment use; (2) allow for research-campaign-running machine-learning agents with diverse learning capabilities and goals; and (3) facilitate multi-agent collaborations and teams. The multi-agent autonomous facilities scalable framework (MULTITASK) makes possible facility-wide simulations, including agent-instrument and agent-agent interactions. Through modularity, real-world facilities can come online in phases, with simulated instruments gradually replaced by real-world instruments. We hope that MULTITASK will open new areas of study in large-scale autonomous and semi-autonomous research campaigns and facilities.

自主材料研究系统使科学家能够更聪明地失败，更快地学习，并在他们的研究中花费更少的资源。随着这些系统在数量、功能和复杂性方面的增长，一个新的挑战出现了——它们将如何在大型设施中协同工作？我们探索一种解决方案——多代理实验室控制框架。该框架是在自主材料科学实验室的基础上进行演示的，可以将来自不同研究活动的信息结合起来解决科学问题。该框架可以 (1) 考虑现实的资源限制，例如设备使用；(2) 允许研究活动运行的机器学习代理具有不同的学习能力和目标；(3) 促进多代理协作和团队。多代理自主设施可扩展框架 (MULTITASK) 使设施范围内的模拟成为可能，包括代理-仪器和代理-代理交互。通过模块化，现实世界的设施可以分阶段上线，模拟仪器逐渐被现实世界的仪器取代。我们希望 MULTITASK 将在大规模自主和半自主研究活动和设施中开辟新的研究领域。