Background
Modern AI systems are evolving from isolated models into networks of collaborating AI agents that work together to achieve complex goals. Examples include autonomous transportation systems, distributed robotics platforms, smart manufacturing environments and large-scale digital assistants.
A key challenge is determining how these agents coordinate their behaviour while competing for limited computing resources across geographically distributed infrastructure.
Research Challenge
The next generation of AI systems will consist of large numbers of specialised agents that collaborate to solve problems that are too complex for a single model. These agent networks may be responsible for coordinating autonomous vehicles, managing smart-city infrastructure, orchestrating robotics fleets, or supporting large-scale AI assistant ecosystems.
A key challenge is enabling agents to make effective decisions while operating across geographically distributed edge infrastructures with limited resources and incomplete information. Agents must coordinate tasks, negotiate access to resources, share knowledge and adapt their behaviour as workloads and environmental conditions change.
This project investigates how Multi-Agent Reinforcement Learning (MARL) and other decentralised AI techniques can be used to orchestrate large-scale agent systems. The research will explore challenges including cooperative decision-making, agent communication, resource allocation, fault tolerance and scalability in dynamic edge environments.
Example application domains include autonomous transportation systems, collaborative robotics, smart energy networks and next-generation AI assistant platforms. The work will involve developing and evaluating intelligent orchestration strategies for large populations of interacting AI agents.
Topics
Multi-Agent AI systems; Cooperative reinforcement learning; Agent communication protocols; Distributed optimisation; Edge-native architectures; Autonomous orchestration
Impact
Many researchers believe that the future of AI lies in agentic systems rather than standalone models. Understanding how large populations of AI agents can coordinate effectively represents a major open challenge in artificial intelligence and distributed systems.