New Software Stops Mars Rover Confusion

MattSparkes writes "The Mars rover Spirit used to get quite confused when it came upon a rock. Because it could only plan routes of a metre or two it couldn't understand how to navigate around large objects, and frequently used to rock back and forth for hours trying to figure it out. NASA have written new software called D* for the rover Opportunity, which should allow it to autonomously plan routes up to 50 metres long. The new software still won't be able to avoid sand-traps, though."

Comments

Based on A*?

[benhocking]

I'm guessing the name is meant to suggest similarities to A*, but you think they'd mention that in the article.

Re:Based on A*?

[Anonymous Coward]

I've implemented D*, and while more efficient than A* replanning from scratch in terms of heap percolates, the overhead of D* bookkeeping actually made it slower than just rerunning a tuned A* implementation, with the caveat that this was on maps that I generated; I'm sure one can generate a pathological terrain that would erase the difference. The massive reduction in heap percolates is interesting from a computer science perspective, but in my experience, in the real world D* doesn't really give you new capabilities that just rerunning A* couldn't do. Map representation e.g. quadtree has a much more drastic effect on performance.

Re:Surprised

[drinkypoo]

Is anybody else amazed at the apparent simplicity of a lot of the problems facing the rover?

Pathing in a video game is (by comparison) easy because you don't have to translate into the real world.

Think about what goes into pathing in the real world for a moment. You have to consider the terrain you're crossing and anything else that might get in your way. You have to be able to revise your model intelligently as you get closer to things and get new sensor data. The whole problem is immensely complex.

In a computer game, the entity pathing already knows precisely where every object is in the map any time it wants to think about it. It knows precisely where it is. It knows exactly how fast it can move at all times. If there even is any difference between various types of terrain, the precise result of being on each type of terrain is known (and thus probably loaded from a precomputed table.)

The only thing easier in this case as compared to, say, pathing in a FPS, is that Mars can be modeled as a heightfield. The rovers are not going to be traveling on bridges any time soon (although that would certainly be news.)

Geology vs. software development

[heroine]

It seems since the first year, the rovers have been more useful as a means of testing robotics software than as a means of studying geology.

Re:Waste of cash?

[Xerotope]

When the problems on this planet make it uninhabitable, it'll be nice to know what Mars is like before we try to move there.

And in comparison to the federal budget for social security and the military (combined something like 85% of the total budget), a few million for a Mars rover is a drop of the bucket.

It's really a question of values, and what you consider to be "problems" that are more important. That's a subjective judgment, and you can't really say it's "wrong" to spend money exploring Mars.

Re:This is pretty impressive....

[jeffeb3]

"My forays into hobby robotics tells me that this isn't as easy as it first appears it might be."

Thank you. I program autonomous vehicles, and people sort of stare at me blankly when I tell them I've spent the last few months of my life trying to detect the color of a traffic light. People often take for granted their ability to sense the world around them. I can assure you, robots do not take that for granted...

Re:Thank you: Why can't New Scientist do this?

[radtea]

Here's one that that really bugs me about online news articles: they rarely take advantage of the medium! Why the hell couldn't New Scientist provide a link to this paper?

Because science journalists and editors know nothing about the fields they report on, and frequently seem actively hostile to the facts in favour of some made-up, mythological story.

Some years ago a colleague at the university where I was a post-doc came into the lab and said, "Today the science section of the LA Times has five stories about stuff that I am either personally familiar with or have actually worked on. Four of them are all but unrecognizable. The fifth is full of things that are not true."

In my subsequent experience as a scientist I have found this to be pretty much par for the course for science journalism. I don't actually know if reporters are as ignorant as they appear, but by the time the edited work gets to the reading public it is almost always spun in such a way as to be misleading or simply wrong.

There seem to be some myths about science that are the bedrock of science reporting, and reporters or editors will distort or simply lie to ensure they reinforce those myths. Some of them are:

1) Science is mysterious and full of contradictions (see yesterdays "hot ice" story)

2) There is no point in anyone trying to understand science, it is beyond you (thus the lack of useful links in stories)

3) Scientists are either put on a pedestal or dragged through the mud. They are never treated as merely ordinary people doing a job or following a vocation.

4) The "story" is more important than the facts.

There are probably a lot more, but basically, science journalism will never be worth anything until it starts actually reporting on science rather than wasting time promulgating editorial myths.