Roadmap
What we are doing and why
Goal
Our goal is to enable robots in the real-world to cooperate in real-time without human intervention.
Why?
There are environments that are dangerous or unstable, where human operators would prefer to remain remote and intervene only intermittently with high-level instructions. We might call these environments deep, eg:
deep sea
deep earth (caves, remote mountains, etc)
deep space
There are also tasks currently done by humans involving heavy vehicles and machinery that demand low-latency decision making and effectively zero probability of failure. Air-Traffic Control is a current example, and it is not hard to imagine situations in the near future involving autonomous vehicles and peer-to-peer traffic control, with similar requirements but no control tower.
Obstacles
People tend to assume this type of cooperation is already possible and just a matter of building the robots and programming them (or training their AI) to cooperate. That is not the case, largely because programming (or training) emergent behaviour is hard and making any hard guarantees about that behaviour is even harder. On top of that there are various problems with generals.
When we say robots coordinating or robots cooperating we mean robots programmed and deployed independently, operating in the same space, communicating to achieve some goal or avoid some disaster.
There are few examples of robots coordinating in real-time in the real-world, and those that exist are highly constrained:
drone-swarm light displays are pre-programmed to perform specific movements and can only adapt to a narrow range of situations
military drones are controlled remotely by human operators
So the first obstacle is software. The solution is real-time asynchronous consensus.
The next obstacle is to deployment. Artificial intelligence sophisticated enough to operate in the real-world is a recent phenomenon and not widely deployed. That limits what we can do since robots must be intelligent enough to operate independently in the real-world before they can cooperate.
There are stepping stones to real-world deployment that do not have this limitation.
Roadmap
1. Shared-memory operating system
This enables us to program many machines as if they were one machine, no matter the machines: robots, vehicles, web servers, etc. Progress on one is largely transferable to the rest.
2. Web applications
We have built several web applications with advantages over legacy applications.
3. Real-time peer-to-peer gaming
This is the natural next step from web applications and it will be a huge step forward for users. Nothing like it exists because it is not possible with legacy technology. Blockchains are too slow and the alternatives are not robust enough.
4. Real-time real-world peer-to-peer applications
The goal.