Methods and Systems for Autonomous Agent Control


Invention Summary:

The challenges posed by high-risk and hazardous environments include knowledge gaps and unknown territories, rendering the deployment of human personnel both dangerous and undesirable. Unmanned Aerial Vehicles (UAVs) have increasingly become the preferred solution in these scenarios and to support Search-and-Rescue (SAR) missions. However, a key challenge for multiple UAVs is maintaining stable communication, despite disconnections, to ensure efficient coordination and control.

Rutgers researchers have developed two novel coordination frameworks––Hierarchical Multi-Agent Actor-Critic (HMAAC) and Transformer-based Communication-Aware Multi-Agent Actor-Critic (TMAAC)––to enable multi-agent UAV control in real-world scenarios. In HMAAC, a high-level policy is deployed at a Coordination Point (CP) to help reduce interdependence among agents. Each agent communicates with the CP to gather information about others. TMAAC enables agents to coordinate in a localized manner, where agents only communicate with their reachable neighbors. HMAAC and TMAAC can be seamlessly combined from one to the other depending on the communication and environment condition. Benchmark evaluations and Microsoft’s AirSim simulations support their efficiency and superiority over state of the art; specifically, TMAAC is shown to be robust even in unreliable communication scenarios.

Market Applications:

  • Search-And-Rescue (SAR) missions in remote or communication-compromised areas.
  • Autonomous surveillance and reconnaissance in unstable environments.
  • Efficient management of UAV fleets for various commercial and military applications.

Advantages:

  • Enhanced adaptability to unreliable communication conditions.
  • Dual-modality execution approach for flexibility in agent operation.
  • Utilization of transformer-based MARL for improved coordination and efficiency.

Publications: •    Sun, C., Huang, S., & Pompili, D. (2023). HMAAC: Hierarchical multi-agent actor-critic for aerial search. IEEE ICRA 2023, 7728-7734.

Intellectual Property & Development Status: PCT application filed, patent pending. Available for licensing and/or research collaboration. For any business development and other collaborative partnerships contact marketingbd@research.rutgers.edu

Patent Information: