In a small and agile team inside GoodAI, we have set out to solve the challenges in drone operations and develop tools for the collaboration of heterogeneous groups of intelligent agents with humans. This work builds on GoodAI’s research history in multi-agent systems.
Here is the “storyboard” introducing our vision, which we’re busy turning into reality:
At GoodAI we’re always looking for use cases where AI can bring the most benefit and truly augment human capabilities. We saw that drones could have been a game changer in safety and security, if only they could be more affordable and robust (in other words, if instead of a few expensive drones one could use a smart swarm of smaller, disposable units), and more accessible (so one could run complex drone operations without special training).
Simplifying Drone Operations
Our first goal is to create a ground control station with an intuitive User Interface, allowing users to plan and oversee routine and rapid response drone missions. The user would only specify high-level goals, allowing the drones to adapt to changing environments and tasks. We’re also exploring natural language communication with drones using advanced AI algorithms like Large Language Models (LLMs) as a way to manage the complexity of multi-drone operations.
Our second goal is to enable drones, and eventually any kind of agents, to collaborate with humans in groups and swarms. We treat heterogeneous drone groups as a robust tool that could self-organize and reconfigure depending on the task at hand. We design our ground control station to support the collaboration of any type of drone, ground robot, or sensor, all through one interface.
The following raw demos give an idea of the general use case and current capabilities we further develop and customize for our clients.
Patrol and detection:
Current Use Case: Community Crime Prevention
Since February 2023, we’ve been looking at safety and security use cases starting with a pilot project with the Community Crime Prevention, a non-profit organization in Hout Bay, South Africa. Community Crime Prevention receives alerts from residents through various channels, including WhatsApp, phone calls, and their GPS-based alert app. The goal is to empower them with a tool to assess situations quicker, cover vast wilderness areas for Search and Rescue, and patrol high-risk areas for crime prevention.
To address these challenges, we’re focusing on:
- Automating drone operations for Search and Rescue, patrolling, and alert/panic button response scenarios, reducing the need for skilled drone operators.
- Implementing a user-friendly Graphical User Interface (GUI) with key operational and safety features, ensuring a fast and easy onboarding process in the control room.
- Developing collaborative scenarios where drones rotate in patrolling and automatic deployment to ensure smooth operations.
- Exploring collaborative scenarios like covering larger areas faster, sharing responsibilities among the drones once they detect anomalies, easing human/robot collaboration through experimental methods like natural language communication, and enhancing robustness.
Our vision goes beyond the immediate use case, offering wide-ranging benefits:
- Extending applicability to other domains requiring quick situational awareness, rapid response, wide area coverage, and robustness
- Streamlining customization of the solution, allowing other developers to tailor the UI and underlying robot control easily.
- Adapting the UI and ground control station to diverse drones and future compatibility with other robots and sensors, including specialized custom-made robots. Ultimately, our focus is multi-agent collaboration, no matter what the agent is.
We would like to see our work help democratize drone technology and augment safety where it’s needed most, by simplifying operations, improving multi-agent capabilities, and enabling real-time adaptability to response scenarios.
We will keep you posted about our progress, and if you would like to become part of our project, reach out to us.