This page outlines multiple projects I have been a part of. Not all projects are listed.

For a definitive list of papers, please visit the Soorati Lab site.

MARLIN

Multi-Agent Reinforcement Learning Guided-by Language-Based Inter-Robot Negotiation (MARLIN) was a project exploring the use of language models to reduce the training time of multi-robot systems. This is currently published as a pre-print on Arxiv and can be found here. We integrated OpenAI GPT LLMs, Google’s Gemini models, and several Llama models. The reinforcement learning code used the Ray library and was eventually deployed on two TurtleBot3 robotic platforms using ROS2 Humble. The GitHub repository can be found here.

SwarmNet

SwarmNet was developed as part of my undergraduate dissertation to create a WiFi-based mesh network between robots in a swarm. This was written in Python and uses a callback-paradigm to react to incoming messages. This project is not currently under development, but alterations and changes are still possible to increase usability. The system uses sockets to communicate and works with both standard WiFi networks and ad-hoc networks. SwarmNet can be downloaded from PyPi. Source code can be found here.

Language-Based Robot Negotiation

My undergraduate dissertation investigated the usage of language for multi-robot coordination. Adjacent to this, we published a demonstration at AAMAS 2024 (Demo Track) highlighting the use of LLMs to enable inter-robot negotiation. The paper can be found here and the demo video is available on YouTube. My dissertation is available upon request.