kvadratinho/mars-avatar-project

The Mars Avatar Project is not a piece of software you can run. It is a conceptual research paper, written and published as a white paper, that proposes a new way of exploring Mars. Instead of sending astronauts to do physical work on the planet, the idea is that people stay on Earth and control a fleet of robotic avatars that provide a physical presence on the Martian surface in their place. The core problem the paper addresses is the communication delay between Earth and Mars, which ranges from about four to twenty two minutes one way. Because of that delay, a human cannot control a robot in real time the way you would a remote-control car. Instead, the paper proposes that human operators set the mission goals, and autonomous systems on the robots figure out how to carry them out locally, checking back with humans only for approval on important decisions. The paper lays out a five-layer reference architecture for the "Mission AI" that would run this system, covering how goals get interpreted, how a fleet of robots gets planned and coordinated, how they share a model of the world, how execution is monitored, and how a safety and approval gate works. It also describes a "Goal Package Protocol," a structured way of writing mission goals that includes the objective, constraints, success criteria, how much autonomy is allowed, and a fallback plan if something goes wrong. A related idea, called Decision Classes, is a scale from decisions that need human approval to decisions the robot can make entirely on its own. The author positions Mars as the hardest possible testing ground for this kind of technology, one that could later apply to the Moon, asteroids, or other deep space locations, and compares the proposal to existing efforts such as NASA AutoNav, NASA AEGIS, ESA METERON, and the ANA Avatar XPRIZE. The repository itself contains the white paper as a PDF and Word document, a presentation, the README, and a license file. There is no code to install or run.