IPB

Welcome Guest ( Log In | Register )

6 Pages V  < 1 2 3 4 5 > »   
Reply to this topicStart new topic
First 2009 MSL Landing Site Workshop
mcaplinger
post Apr 23 2006, 04:10 PM
Post #31


Senior Member
****

Group: Members
Posts: 1293
Joined: 13-September 05
Member No.: 497



QUOTE (Phil Stooke @ Apr 23 2006, 08:24 AM) *
We do get mixed messages about daily and total range, probably from sources written at different times as their thinking evolves.

Definitely true. And one has to consider the source too; there appears to be little consensus yet about what MSL operations will actually look like.

The current engineering constraints document says this:

"As part of its primary mission, the MSL rover would include the capability for traversing long distances. Currently, the system is being designed for a total actual traverse distance capability of no less than 20 km. For purposes of hardware life and cycle evaluation, it is assumed that this traverse occurs over a terrain with an average rock abundance of 15%, an average slope of 5 degrees, and an average slip rate of 10%. Under these conditions the rover would travel on average about 100-150 m/sol."

But Phil is probably correct that a lot of time will be spent at a particular site before moving on to the next site (as I understand it it takes a fair bit of time for SAM and Chemin to do their things), so the per-sol average traverse is probably not representative.


--------------------
Disclaimer: This post is based on public information only. Any opinions are my own.
Go to the top of the page
 
+Quote Post
Guest_Richard Trigaux_*
post Apr 23 2006, 05:05 PM
Post #32





Guests






QUOTE (Cugel @ Apr 23 2006, 02:19 PM) *
(snip)
Besides being nuclear powered I think the greatest performance increase must come from software development. With all the lessons from MER under the belt it must be possible to build really better autonomous driving programs. I would think.



QUOTE (Phil Stooke @ Apr 23 2006, 03:24 PM) *
To my mind the key to range will be the nature of the software governing driving. If the site has a fair bit of relief, as I would expect, the ability to plan drives over a few hundred metres will be limited even with MRO data. A longer drive in one day will require automated hazard avoidance capability. But if you detect a hazard, what do you do? If you stop and wait for instructions, driving will be slow. If you can try multiple paths until a safe route is found to the designated target, driving longer distances in a day is more feasible. For instance, we might imaging the planners giving instructions to follow a pre-planned route, but offering alternative routes to the same place based on MRO data. If MSL is stopped by an unexpected hazard, it could search locally for a way round the hazard, or retrace its steps to a branch point and follow the second alternative route, without intervention from the ground. That would be faster. But I don't know anything about the strategy to be followed on MSL.

Phil



I agree with both of you. Autonomous capability is the key for long range and several targets.

I would add that they should design the rovers with a long lifetime and long range. It would add a bit of weight (for instance the only way to increase the lifetime of a ball bearing is to increase its size) but this extra cost will be recovered with less launchs. Provided of course that there are enough science targets within range.
Go to the top of the page
 
+Quote Post
mcaplinger
post Apr 23 2006, 05:19 PM
Post #33


Senior Member
****

Group: Members
Posts: 1293
Joined: 13-September 05
Member No.: 497



QUOTE (Richard Trigaux @ Apr 23 2006, 10:05 AM) *
I agree with both of you. Autonomous capability is the key for long range and several targets.

I agree too, for suitable definitions of "autonomous", "long", and "several". But for autonomy as JPL has implemented it on MER, and for the range of MER, the autonomy is not frequently used because it doesn't work well enough to do anything but the simplest tasks. Certainly there's no high-level route planning like the sort Phil mentioned. And I'd be surprised if MSL will have or need any more autonomy than MER given the relatively small traverse distance they're talking about.

Now, we're currently working on a much smaller rover with much simpler autonomy that would have longer range (see http://www.amerobotics.ou.edu/research/sr2/ ), but I don't believe JPL is thinking along those lines.


--------------------
Disclaimer: This post is based on public information only. Any opinions are my own.
Go to the top of the page
 
+Quote Post
Guest_Richard Trigaux_*
post Apr 23 2006, 06:13 PM
Post #34





Guests






QUOTE (mcaplinger @ Apr 23 2006, 05:19 PM) *
And I'd be surprised if MSL will have or need any more autonomy than MER given the relatively small traverse distance they're talking about.


software has no weight! The only harware which could add autonomy are a LIDAR (to have a true 3D scenery without using error-prone stereo vision) and a better aaptative wheel suspension. Otherwise to add "intelligence" into the software (such as alternate routes) adds cost only in the development stage. This cost is well repaid at time of roving on bad terrain, finding unexpected target, or getting the rover out of a dune or loose rocks.
Go to the top of the page
 
+Quote Post
mcaplinger
post Apr 23 2006, 06:40 PM
Post #35


Senior Member
****

Group: Members
Posts: 1293
Joined: 13-September 05
Member No.: 497



QUOTE (Richard Trigaux @ Apr 23 2006, 11:13 AM) *
software has no weight! ... Otherwise to add "intelligence" into the software (such as alternate routes) adds cost only in the development stage.

Have you ever tried to write autonomy software? Regardless of what the typical AI/robotics researcher will claim, reliable high-level autonomy of the sort being discussed is well beyond the state of the art.

You might want to read about the work of Rod Brooks -- http://www.kk.org/outofcontrol/contents.php -- for a discussion of the sort of autonomy we're working on.


--------------------
Disclaimer: This post is based on public information only. Any opinions are my own.
Go to the top of the page
 
+Quote Post
Cugel
post Apr 23 2006, 07:25 PM
Post #36


Member
***

Group: Members
Posts: 153
Joined: 11-December 04
Member No.: 120



QUOTE (mcaplinger @ Apr 23 2006, 06:40 PM) *
Have you ever tried to write autonomy software? Regardless of what the typical AI/robotics researcher will claim, reliable high-level autonomy of the sort being discussed is well beyond the state of the art.


Funny that you mention it. It's how I make a living! Unfortunately, I must agree with you. We are far from completely autonomous rover navigation software. But some things speak in our favor:

1. As being a relatively young field of interest, really substantial progress is possible within the next few years.
2. Our militairy friends are very interested in this, as more and more of their systems are becoming unmanned.
And because they are the ones with the big budgets and other resources we can hope to see some serious results. (A lot of people are currently working on this, it's big business!)
3. Computer hardware and sensor development is also on our side (Moore's law).
4. We don't need completely autonomous navigation right now, an incremental improvement would be fine and give us much.

Personally, I think with MSL it should be possible to do 1 km. drives completely autonomously (taking a week or so without further commands) in a Meridiani like environment. No science target selection, of course. Just obstacle detection and avoidance.
Go to the top of the page
 
+Quote Post
J.J.
post Apr 23 2006, 07:30 PM
Post #37


Junior Member
**

Group: Members
Posts: 85
Joined: 22-March 06
Member No.: 722



I'm personally jonesing for Eberswalde Delta...I think the science in the delta proper, the crater floor, and the crater walls (presumably an old shoreline) would be fantastic.


--------------------
Mayor: Er, Master Betty, what is the Evil Council's plan?

Master Betty: Nyah. Haha. It is EVIL, it is so EVIL. It is a bad, bad plan, which will hurt many... people... who are good. I think it's great that it's so bad.

-Kung Pow: Enter the Fist
Go to the top of the page
 
+Quote Post
RNeuhaus
post Apr 23 2006, 07:43 PM
Post #38


Senior Member
****

Group: Members
Posts: 1636
Joined: 9-May 05
From: Lima, Peru
Member No.: 385



QUOTE (mcaplinger @ Apr 23 2006, 12:19 PM) *
But for autonomy as JPL has implemented it on MER, and for the range of MER, the autonomy is not frequently used because it doesn't work well enough to do anything but the simplest tasks. Certainly there's no high-level route planning like the sort Phil mentioned.

I think that the one of the factors that limited the MER's autonomy is due to the fact that their surface image from MGS and/or Odyssey does not provide enough surface resolution as the MRO will provide to MSL. On the other hand, one of the most important keys for the longer autonomy is due to the low height of the MER's mast PANCAM (1.4 m) that hinder the capability to anticipate better the forward surface conditions.

Rodolfo

QUOTE (Richard Trigaux @ Apr 23 2006, 01:13 PM) *
Otherwise to add "intelligence" into the software (such as alternate routes) adds cost only in the development stage. This cost is well repaid at time of roving on bad terrain, finding unexpected target, or getting the rover out of a dune or loose rocks.

It is a very complicated matter to develop a high intelligence software to detect the subtle difference among the different kind of sand (loose and compact), besides takes the degree of slope, temperature of sand (hot sand, is sleeper since it is drier, wet sand, is better for traction) and also the relieve of sand also takes into the account on how to maximize the traction. On the other matter, about the safety, to decide to go whenever every aspects are meet the safety rules such as the stone height, slope, loose surface, wet surface, slippery surface and even inclusive any bigger saps that might appear than ones of often of Meridiani.

On the overall, I think we still have very little experience about the tricks of Mars surface. So the MSL must continue driving with care since it is still one of the space pioneers!

Rodolfo
Go to the top of the page
 
+Quote Post
mcaplinger
post Apr 23 2006, 07:56 PM
Post #39


Senior Member
****

Group: Members
Posts: 1293
Joined: 13-September 05
Member No.: 497



QUOTE (Cugel @ Apr 23 2006, 12:25 PM) *
Our militairy friends are very interested in this, as more and more of their systems are becoming unmanned.

Well, maybe; of course, AI researchers have been saying exactly the same thing since the early '80s. The DARPA Grand Challenge ( http://www.darpa.mil/grandchallenge/ ) does lend some credence to the state of the art, but while software may weigh nothing, the hardware and sensors that drove the Grand Challenge vehicles were orders of magnitude heavier than what we could put on a Mars rover. (Memory and CPU cycles to run that software at usable speeds certainly weighs something!) Also, I think all the GC vehicles had to do was stay on the road, which is a totally different problem than that faced by a Mars rover.

Another point: most unmanned military vehicles, UCAVs for example, are teleoperated, not autonomous. The exceptions are those that fly simple courses with no sensors, like Global Hawk.

I stand by my original statement: we are nowhere close to deploying useful high-level autonomy on Mars.


--------------------
Disclaimer: This post is based on public information only. Any opinions are my own.
Go to the top of the page
 
+Quote Post
Guest_Richard Trigaux_*
post Apr 23 2006, 08:19 PM
Post #40





Guests






QUOTE (mcaplinger @ Apr 23 2006, 07:56 PM) *
(snip)

I stand by my original statement: we are nowhere close to deploying useful high-level autonomy on Mars.


Yes, certainly. But some could be gained, relative to Oppy and Spirit, which could make a difference. More autonomy should certainly be possible without adding much mass, and within reach of the team who will actually write the programs. For instance trying alternative route when one don't work requires just adding lines of code, and it could save a day of roving at some occasions. Or having a fan of acceptable paths. The road is found by try and mistakes, actual computers are good for this.
Go to the top of the page
 
+Quote Post
Jeff7
post Apr 23 2006, 08:27 PM
Post #41


Member
***

Group: Members
Posts: 477
Joined: 2-March 05
Member No.: 180



QUOTE (Richard Trigaux @ Apr 23 2006, 02:13 PM) *
software has no weight! The only harware which could add autonomy are a LIDAR (to have a true 3D scenery without using error-prone stereo vision) and a better aaptative wheel suspension. Otherwise to add "intelligence" into the software (such as alternate routes) adds cost only in the development stage. This cost is well repaid at time of roving on bad terrain, finding unexpected target, or getting the rover out of a dune or loose rocks.


True, but a computer powerful enough to run that software at a respectable speed does have weight. The MER's only use a 20MHz processor.
A more powerful computer may be slightly larger, and it will require more power.


And the other problem is, making a computer that has some degree of intelligence is difficult. Computers are stupid. Incredibly, unbelievably stupid. They will do only what they are told, nothing more, nothing less. If a rover is told to drive over a cliff, or into an obstruction that will damage it, it will dutifully do it, unless a programmer early on told it specifically not to do that. And if the programmer tells it "Don't drive into obstructions that slope 50 degrees to the left," and you try to drive the rover into something sloping the other direction, it'll smack right into it without a second thought.

That's just basic object avoidance. Having a computer then figure out the "best" route to get somewhere requires a whole new level of complex programming. Having it find multiple routes yet, and evaluate multiple factors, such as traversability and distance travelled, and then choose, that too is difficult.

Add to all of that that you want some kind of fault mode programming buried in there for when something goes wrong, like what happened with Spirit's flash memory problem. Fortunately, it had a fault mode that allowed for very low-level functionality.

QUOTE
Other sources talk about 10 km. during its lifetime of 2 Earth years.

That just amused me briefly - the MER's are already around 7km each. smile.gif
But I get MSL would catch up to that mark very quickly.
Go to the top of the page
 
+Quote Post
Phil Stooke
post Apr 23 2006, 08:47 PM
Post #42


Senior Member
****

Group: Members
Posts: 5819
Joined: 5-April 05
From: Canada
Member No.: 227



I should probably make it clear that I am aware of the difficulty of programming autonomy into the driving.

In answer to one point, I wasn't thinking that the rover would do its own route selection, or multi-route planning, but that those things would be done on earth before each drive. The rover would follow a plan until an obstacle stopped it - and of course the plan would already avoid most of them. But once you are stopped by an unexpected obstacle, instead of stopping right there, you would ideally be able to back up one or two metres, check for safety to the left and right, and take a detour to try to get around the obstacle and back on the pre-planned track. Spirit does this now, from time to time - there have been instances well recorded in images of the tracks, for instance at Arad on 716 and just coming down off Home Plate on 779. I think a bit of improvement in that line is not unreasonable and would be very productive.

Phil


--------------------
... because the Solar System ain't gonna map itself.
Go to the top of the page
 
+Quote Post
The Messenger
post Apr 23 2006, 09:04 PM
Post #43


Member
***

Group: Members
Posts: 624
Joined: 10-August 05
Member No.: 460



QUOTE (AlexBlackwell @ Jan 18 2006, 09:57 AM) *
The workshop will be held in the vicinity of JPL in Pasadena, CA, and there will not be a registration fee. In order to get a sense of the number of people likely to attend the workshop, interested individuals should indicate their intent to attend via http://marsoweb.nas.nasa.gov/landingsites/ by April 1st, 2006. Although we anticipate mostly oral presentations, there may also be poster sessions. Additional logistical information about the workshop will be distributed to the community in subsequent announcements and will be posted at: http://marsoweb.nas.nasa.gov/landingsites/ and http://webgis.wr.usgs.gov/msl Input from the science community is critical to identification of optimal landing sites for the MSL. We look forward to your involvement in these activities!

Regards,

John Grant and Matt Golombek
Co-Chairs, Mars Landing Site Steering Committee

I notice in the mission planning, they recommend in situ measurements of temperature, density and pressure during descent at a sampling rate of 100 Hz. I think that with current techology, this could be quaddrupled with no weight penalty, and I would add a least three axis of acceleration to the mix, and an attempt should be made to telemeter all of this data in real time.

After seeing the dynamic buffeting both MER's suffered, the more data during this phase, the better.
Go to the top of the page
 
+Quote Post
Guest_BruceMoomaw_*
post Apr 24 2006, 01:01 AM
Post #44





Guests






Back in June 2002, the MSL science steering group wanted MSL to have substantial onboard autonomy allowing it to drive between its locations for detailed sampling and study at the rate of fully 450 meters/sol, or 3 km in 13 sols -- and, once it arrived, to "be able to approach a designated target, deploy the instrument arm, mini-corer, or drill, and begin science activities (measurements or drilling/coring), using only a single command cycle to initiate the full suite of activity." ( http://trs-new.jpl.nasa.gov/dspace/bitstre...4/1/02-1822.pdf ) . A typical plan based on this idea involved it, during 667 sols of operation, driving about 69 km to study 23 different detailed study locations for about 16.5 sols each.

Unfortunately, that plan then went a-glimmering, and they went back to the MER level of driving and target-approach autonomy ( http://trs-new.jpl.nasa.gov/dspace/bitstre...7/1/03-2974.pdf ; http://www.ninfinger.org/~sven/models/vaul...121secure31.pdf ), in which it would traverses only about 50 meters/sol and take a 3-sol cycle to approach any particular sampling target at one of its detailed study locations. For the reasons, see the first of those two documents, pg. 147:

"(1) PSlG wanted to balance science and engineering sophistication: Mission life driven much less by driving range, speed or hazard detection autonomy than by number of science decisions requiring human interaction at a rock sample site.

"(2) Large vehicle size allows for simple path planning.

"(3) Consistent with an 'autonomy to cost' strategy:
Hazard detection and avoidance test cost could be unbounded.
Could infuse more autonomy once science objectives are met."
______________________________


Assuming, again, that MSL spent a total of about 381 sols at its detailed study locations, it could drive only about 1/9 as far as in the earlier plan -- that is, about 7.7 km. And the new plan called for it to acquire 74 samples, each of which would now take about 7 sols to approach and collect -- so we're talking only about 150 sols worth of driving at 50 meters/sol, which again came out to only about 7.5 km.

Well. Now we're back up to an ability to traverse 100-150 meters/sol during long drives, so -- assuming that we still plan to spend about 380 sols collecting samples -- we are indeed back up to 15 to 22.5 km total drive distance.
Go to the top of the page
 
+Quote Post
Stephen
post Apr 24 2006, 02:17 AM
Post #45


Member
***

Group: Members
Posts: 307
Joined: 16-March 05
Member No.: 198



QUOTE (Cugel @ Apr 23 2006, 02:19 PM) *
http://www.nuclearspace.com/a_2009_Rover.htm
This article talks about 'miles'.

http://space.com/businesstechnology/060118_msl_wheels.html
Is talking about 'hundreds of meters per day'.

Other sources talk about 10 km. during its lifetime of 2 Earth years.

Emily's "Report from MEPAG" entry on her blog reports Bruce Betts telling her that the MSL "will have a nominal mission distance of at least 20 kilometers".

======
Stephen
Go to the top of the page
 
+Quote Post

6 Pages V  < 1 2 3 4 5 > » 
Reply to this topicStart new topic

 



RSS Lo-Fi Version Time is now: 27th November 2014 - 05:12 PM
RULES AND GUIDELINES
Please read the Forum Rules and Guidelines before posting.

IMAGE COPYRIGHT
Images posted on UnmannedSpaceflight.com may be copyrighted. Do not reproduce without permission. Read here for further information on space images and copyright.

OPINIONS AND MODERATION
Opinions expressed on UnmannedSpaceflight.com are those of the individual posters and do not necessarily reflect the opinions of UnmannedSpaceflight.com or The Planetary Society. The all-volunteer UnmannedSpaceflight.com moderation team is wholly independent of The Planetary Society. The Planetary Society has no influence over decisions made by the UnmannedSpaceflight.com moderators.
SUPPORT THE FORUM
Unmannedspaceflight.com is a project of the Planetary Society and is funded by donations from visitors and members. Help keep this forum up and running by contributing here.