Machines, not Meat, to the Moon

MoonMachines

I'm keen to adapt how we think about space to the realities and opportunities of the 21st century. Putting more astronauts on the moon seems pointless; using the same resources, and modern technology, we can put a thousand virtual researchers on the Moon with robotics and terabit laser telemetry.

We do not have the capabilities that put Apollo on the moon, nor do we have the geopolitical justification. Before Neil Armstrong set foot on the moon, a cash-strapped NASA had shut down the Saturn V assembly lines; their political purposes had been achieved. We got the glory, the Soviets got infrastructure that is still usable today.

Nonetheless, we live in exciting times, The Curiosity landing demonstrated what modern technology can do in space today, and on the battlefield soon (sadly). Among many achievements, Curiosity landed within 1.5 miles of target center, three times as accurate as Apollo 11 after travelling 1000 times the distance. Modern space missions succeed because they replicate physics-accurate missions that have already occured countless times inside supercomputers; add humans, and accurate simulation becomes impossible.

We do not need to send such large, autonomous machines to the moon. We can flashlight-sized remote controlled tools, connected through predictive-adaptive telepresence to trained operators and researchers on earth. No team of ten mission specialists will ever be smart enough or have enough time to match all those researchers - indeed, they will be so busy surviving that they may not have as much research time as one researcher on earth. Would you rather stand in a crowd waving goodbye to ten brave astronauts - or bicycle to work, put on a haptic suit, and spend the day manipulating lunar materials 400,000 kilometers away, looking for world-changing discoveries?

The moon is dead. The rocks aren't going anywhere, there is no changing weather or plants or animals. Few unpredictable events will occur during the three second speed-of-light round trip. With good physics models and abundant powerful computation (server sky!), dropping a rock will have the same same simulated results as it will three seconds later when the telemetry arrives. If it does behave surprisingly differently, we will discover something very important and probably very profitable.

We explore space to find surprises, which lead to scientific advances. Surprise can be antithetical to human survival; we learned far less about the moon using astronauts than we can today using expendable machines. We can send LOTS of machines, design them for perfection, then plan to lose half or more due to the unexpected. We should avidly seek the unexpected, the bizarre, the dangerous; take risks that we would never take with human lives in the balance. Astronauts are hostages, not pioneers. Failure must be an option, as it was during the age of exploration. Many proto-Polynesians died before they learned the ways of the Pacific.

Transistors are a billion times cheaper than they were in 1969. While we are far from the artificial intelligence necessary to replace a human brain, that brain comes in a very fragile package, quite incompatible with the space environment. Electronics may be stupid, but it is far more reprogrammable than the human mind, and it will tolerate much worse conditions. There are many places on earth where brainless bacteria can survive, and humans cannot. We send our robots there.

Space will still be very much a human activity - but discovery requires improvisation, not a rehearsed performance from a checklist. Human improvisation is creative, adaptive, and unpredictable, exemplified by the Skylab rescue mission of Nay 1973, when Conrad, Kerwin, and Weitz travelled to the crippled space station with a rapidly cobbled-together toolkit, and improvised when they got there. Robots, controlled by humans on earth, can now do the same. A moon base operated by remote-controlled robots will have many failures - though none so perilous as a wounded astronaut three days and 10 gees from a modern hospital. We can turn off a damaged robot and ship up exotic repair parts on the next supply mission; we can keep most spare parts on site.

Someday, when we drive launch costs below $10/kg, and it is affordable to launch and maintain thousands of tons of infrastructure per person, we can recreate self-supporting human-compatible environments far from mission control. These will be robot-rich artificial ecosystems, managing tons of biomass, air, water, and other supports. The best way to develop that capability is to create opportunities that generate vast wealth per launched kilogram at very large scale, and drive down the cost of launch through automation and volume pricing.

Server sky will make this happen. When we can deliver gigabit internet to millions of habitations in mid-ocean, we can colonize our seas the way we will later colonize space. We must broaden our capabilities, adaptively and productively, and expand our imaginations to match.

Scaling Lunar Robots

An easy way to make "moon response" mimic "earth experience" is to match kinesthetic responses. Drop an object from a 1.5 meter shoulder height on earth, and it will fall to the ground in 550 milliseconds. On the 1/6th gee Moon, a drop from 0.25 meter height will fall in the same time. So - build robots at 1/6 scale, move them at 1/6 of the speed, etc.

The mass - on Earth - will be slightly increased by the weight of a full immersion haptic suit for the "telenaut". Operators will vary in size and mass - they either adapt to the "standard", or multiple sizes of robots become available. Much depends on how adaptive operators can be, but less training is cheaper.

Robot "metabolism" will be different - they will use power for locomotion and data processing like humans, but they will use less power for "metabolism" and far more for telemetry links. Given the small power-to-mass ratio, a given mass of battery will sustain lunar-scaled robot locomotion 36 times longer than a non-scaled robot, but it will provide far less power for radio links, so relay antennas must be closer, perhaps by a factor of L^1.5 to prolong endurance.

Robots will operate in a field of relay antennas, connecting them to nearby computation servers and laser relays to the server sky constellation in MEO, and from there to human operators on the ground. Teams on the moon can be operated by individual human operators scattered throughout the globe, with a bias towards the world's least expensive scientists and field workers, operating from the rising nations of the world.

Robots do not need general purpose "eyes". Constructing the view that human operators see requires knowledge in terrestrial servers, and superb 3D modelling.

Communication systems should send changes, not unsurprising and redundant information. The telemetry from the robots, to the earth, and for scientific analysis need not be 60 frame-per-second marginally-changing fully-populated images; instead, the local system constructs 3D surface models of the mostly static environment and embellishes them with new information. A robot camera is told what it "should" be seeing, and transmits back only the differences from expectation - which, if it is actively exploring and probing the immediate environment, will be many many differences, but not whole frames.

Exploring 100 square kilometers to 1 millimeter and 32 bit resolution is 40 terabytes; a diffraction limited communication link with 1 meter diameter optical dishes (transmit and receive) sending 1 bit with 10 received 500 nanometer (2.5 eV) photons must transmit around 1 million photons per bit; with 1% (T&R combined) quantum efficiency for the link, each bit requires 250 MeV of transmit energy, so the total energy to transmit a 320 Tbit map is 8e22 eV, about 13 kilojoules. A 40 milliwatt, 1 Gbps link could transmit that much data in half a lunar day. per-robot telemetry may not be a big power problem.