[Zero]

Perseverance landed safely on Mars a few minutes ago, congratulations to Nasa teams!!!
 

[WriterOfMinds]

It uses some machine vision for landing: https://mars.nasa.gov/mars2020/mission/technology/#Terrain-Relative-Navigation

[infurl]

One thing that I recently learned to my horror was that there are some quite severe limitations on the amount of computer power that you can put in one of those rovers. I had always assumed that the limiting factor would be the amount of power that it could generate, but the limiting factor is the hard radiation bombarding the surface of the planet and potentially corrupting the machine's memory. The transistors have to be large and so the processors are roughly equivalent to the devices that we had back in the 90s.

I've spent the day wondering how they managed to get the equivalent of a 386 to do machine vision fast enough for that landing process.

Apparently even on the surface of the Earth, protected within our magnetic field, you can expect on average to get one bit flipped per gigabyte per month, so ECC memory is becoming more important, even on desktop machines. Maybe some of those mysterious crashes that you have experienced over the years really were caused by cosmic rays.

[WriterOfMinds]

I've spent the day wondering how they managed to get the equivalent of a 386 to do machine vision fast enough for that landing process.

If they only needed the system for the landing, I imagine that relaxed the radiation requirements somewhat. They wouldn't have had to plan for total ionizing dose over the entire lifetime of the mission, and the probability of a single-event effect during the 10 minutes or so that the system is operational would be comparatively small. Mitigations like ECC and redundancy could take care of the rest.

A custom hardware co-processor doing some of the heavy lifting is also a possibility.

[HS]

Excuse me if I hypothesize while you two have all the knowledge, but that does seem like robustly a solvable problem. I can imagine multiples of the same program in parallel, they could periodically update to the state of the majority. Or a single program with a redundant dataset, with each unit of information stored as a group of bits, allowing for near-indefinite self repair by periodically switching all bits to the state of the majority, assuming there isn’t any physical damage.

[Zero]

The landing phase is impressive indeed, with the machine "looking to" the ground , recognizing terrain features to figure out where she is.

But if I understand correctly, once on the ground, there's also some AI involved in the autonomous pathfinding process. Well the robot is not in a hurry when looking for the safest path, so I guess it can use (relatively) slow hardware for the task.

We have neural nets in the sky guys!   

[ivan.moony]

It would be cool if they could replicate over there.

[Zero]

Yes, and also grow some potatoes like Mark Watney! though in order to do it Watney-style, the next rover would need to take a very special cargo to Mars

[WriterOfMinds]

Excuse me if I hypothesize while you two have all the knowledge, but that does seem like robustly a solvable problem. I can imagine multiples of the same program in parallel, they could periodically update to the state of the majority. Or a single program with a redundant dataset, with each unit of information stored as a group of bits, allowing for near-indefinite self repair by periodically switching all bits to the state of the majority, assuming there isn’t any physical damage.

Redundancy-and-repair is certainly one of the standard tools in the radiation survival kit, but it comes at a cost. Any redundant circuitry you add increases size, weight, and power consumption, and if you're building a system for space, all of those will probably come at a premium. It's a tradeoff ... you don't want to expend so much effort shoring up fast, fragile hardware that you're worse off than if you had used slow, durable hardware.

Redundancy can even be a double-edged sword -- yes, you have multiple copies in case one takes a hit, but you've also increased the system's active volume, thereby increasing the likelihood that something takes a hit! This is a bigger problem if you leave out the repair system (or if it does not repair fast enough for your radiation environment) -- errors can accumulate until bad copies overrule good ones.

[Zero]

Maybe a continuous stream of healthy data from earth can, in some cases, remind the machine what "truth" is, so to speak? The idea would be to keep the probability of errors low, by refreshing data that's already known by the rover/satellite... some kind of remote backup.

[Zero]

https://youtube.com/watch?v=4czjS9h4Fpg

[Zero]

So... the sky is almost blue, did you notice? We're lucky to live this. What an era!

[infurl]

I was intrigued by how slowly the heat shield seemed to be falling away from the lander when it was released, compared to what I am used to seeing in skydiving videos on Earth. At first I thought it was because of the lower gravity, but then I realized it was because of the very thin atmosphere. The shield and the lander were falling at almost the same very high speed, even with the parachute of the lander deployed, because there wasn't enough drag to slow the lander down much. It really reinforces just how difficult it is to land safely on Mars. Larger planets like Earth and Venus have thicker atmospheres which compensate for the higher gravity and smaller planet(oid)s like the Moon and Mercury have much less gravity so the lack of aerobraking is much less significant.

[Zero]

Apparently it landed within 5 meters of the planned target. One thing I'm waiting for is, they say Perseverance might watch the next rover landing, which means we would have a video of this from an external point of view. And since it has a 3D cam, it would probably feel like you're up there, sitting on a rock and enjoying the show. If you add the sound, this is a promise of a spectacular moment for all earthian space nerds!

[infurl]

Unless Perseverance actually finds something I would be very surprised if they sent another rover to the same place. Given the effort involved they would get a lot more data by sending it somewhere else. On the other hand, the life tests on the original Viking lander spacecraft registered positive results and they have been quietly ignored all these years.

https://phys.org/news/2016-10-year-old-viking-life-mars.html

This was parodied rather brilliantly in the movie "Planet 51"