Swarm Artificial Intelligence – Short Presentation
Short presentation explaining swarm artificial intelligence and its applications in drone swarms and robotics.
WatchOverview
This presentation is about a form of artificial intelligence inspired by swarm intelligence in biological systems.
This is the short version of the presentation (approximately 40 minutes). A full version is also available (approximately 90 minutes):
Swarm Artificial Intelligence (full version)
For best viewing of the computer simulations, use a large screen such as a laptop or tablet.
Different Language Versions
The narration is generated by an AI system trained on my voice, which also translates the speech into other languages. This AI produces a good likeness of my voice in English, but the similarity is less good in some other languages.
Different languages for the narration and text can be selected using the dropdown language selector at the top of the page.
Contents
This version has five parts:
- What is swarm intelligence?
- Why is this important for robotics and drones?
- Creating artificial swarm intelligence
- Simulation examples of swarm AI
- Summary and conclusions
Swarm AI
This is the name I have given to the AI technology I have developed, which has applications in robotic systems and drone swarm warfare.
This type of AI is inspired by a different branch of the tree of life from that of humans. Its key principle is cooperation between many simple agents, rather than a single complex system. I will explain how correctly designed simple systems can produce complex collective behaviour. This effect is both surprising and powerful.
One important point is that this AI does not simply copy natural swarming behaviour. It develops its own group tactics to achieve its goals. In some situations it forms a tightly coordinated swarm, but in others it becomes highly dispersed, depending on the objective and constraints.
Computer Simulations
These include a grid search task, a dynamic communications repeater chain, and several autonomous swarm attacks.
The attack scenarios include a prolonged attack against a point target, a distributed thermobaric attack over a wider area, and an attack against multiple moving ground targets. In all simulations, the objectives are successfully achieved, and all targets are destroyed.