This invention describes a way to manage multiple groups of automated machines, like robots or drones, that are all working towards the same goal. Each group uses an identical artificial intelligence (AI) program, called a neural network, to guide its actions. If one group starts performing better than others, its successful AI program is copied to the other groups, helping all machines improve their performance. The claims specifically detail controlling a first and second group, and the ability to update the AI program on machines while they are still operating, with each machine storing both AI programs.
Why it matters: Filed before the explosion of distributed AI and reinforcement learning techniques, which have significantly matured since 2020. The ability to efficiently train and deploy adaptive neural networks across multiple agents, and the increasing power of edge computing for automated machines, makes this approach more viable and performant today.
AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.
Reinvent this with AI