The stereo visual feature-recognition system seems to work very well for the MERs; my main question is how well it would work on a much faster-moving vehicle of the sort we're talking about. A lidar system capable of automatically detecting dangerous obstacles or slopes ahead and slamming on the brakes would seem to be a useful addition to a manned teleoperator for a lunar vehicle.

As I said there are fundamental illusions that a stereo-vision system cannot dissipate. With the MERs it worked, because the rover looks, create a map, and advance of some tens of centimetres maximum, then stops and redo again. If an illusion is not eliminated at a given turn, it will necessarily be at one of the following. On a fast moving rover this will not be possible, as the braking distance can be greater than the safe vision distance. It is why your idea of a laser telemetre is interesting. It would not only be an emergency monitoring, but a telemetry laser mapping the soil in 3D provides a much safer map than just stereo vision, where hidden hollows are clearly visible and not confused with flat or uniform surfaces. So the 3D map would be interesting even for human operators, in remote control or even in the rover itself.

The biggest pitfall for an automated fast-speed (30-kph-plus) lunar rover is... pitfalls.

Even on the flattest lunar terrains, crater saturation means that there are literally thousands of craters of all sizes along any given straight-line path of even a few thousand meters in length. Lidar could "see" upcoming craters, sure -- as long as the near rim isn't higher than the far rim.

Problem is, there are a *lot* of lunar craters of rover-eating size that have uneven rims. They tend to be completely invisible to *any* line-of-sight system until it's too late to brake to avoid driving straight into them.

The LRVs had a maximum level-ground speed of about 15 kph, and the Apollo experience was that this was about the fastest safe speed to travel, to allow for braking or violent steering when such a hidden crater popped up right in your path. The only time the LRV drivers exceeded that speed was going down hills, where the terrain was spread out in front of them and therefore displayed its pitfalls to easy view.

And that was with human drivers, who were able to react with human reflexes. Robots are still incapable of reacting to driving conditions with anything approaching human reflex speeds. I doubt that's going to change any time in the near future.

Miss just *one* blocky-floored 5- to 10-meter crater and sail right over its rim, and you've just wiped out your billion-dollar robotic explorer.

I think robotic lunar explorers are going to be limited to at most a 10-kph pace for a while, yet...

-the other Doug

Yes, this is a real problem. And a large lunar rover must be fast, otherwise it remains confined in one place. By the way, a rover at 15km/h needs one month to circle the Moon equator, (and at this speed it follows the Sun so that it always get energy). So after all there is no need to be much faster.

But even at 15km/h the problems are huge, and as it, no robot can do it, and no remote controling human operator either, with the 2s delay. Only a manned vehicule can drive in these conditions, and even at its own risks.

Then Bruce's idea of using a LIDAR is interesting. Such lidar beam could scan the landscape and do of it directly a 3D map, without the ambiguities of a stereo algorithm. For instance, in the (very common) case where there is an invisible pit (near rim higher than the far rim) a stereo vision can see only an uniform grey span, and not the pit, until it is too late. But a lidar scanning this place, say from bottom to top, detects an instant increase in distance when the beam leaves the near rim and reaches the far rim. From this it can infer the presence of an unseen area, potentially a pit. And this unseen area could be displayed on a screen as a black span: the operator just has to avoid it. This is why the lidar scan is infinitely superior to the stereo algorithm and even to human vision.


In the long run, a fast rover on the Moon is ubiquitous. Just chinese taikonauts land by surprise? The rover can be here and wave a free Tibet flag... err provide its science analysis facilities.


Note that if the sciences instruments are in racks, and the racks opening to the top, we can conceive an automated lander with big legs, the rover sneaks under, and the lander can automatically change an instrument. Difficult, but not absurd: the task can be resumed to grasping a knob, pull it, and plug another instrument. Basic topic of robotic studies.

Look. Even on Earth, recreational 4WD operators are consistently told NOT to drive fast on rugged terrain, but to "crawl" over obstacles instead, at no more than 10-20 km/hour. Given the radio time lag on a teleoperated rover, cutting the speed to only a few km/hour seems entirely logical. Nor, given the geological-survey purpose of such a vehicle, is there any scientific benefit in driving faster -- you'll just miss interesting sampling sites, which is after all your purpose rather than trying to set a speed record.

http://www.darpa.mil/grandchallenge/
131 miles in 6hrs 53 mins

Call it 18mph, or 28kph.

Yes - I know - not entirely applicable to lunar exploration - but it's untrue to say 'no robot can do it'.

The technology is beginning to be available whereby it's entirely plausable that a rover could operate on the moon at a pace similar to that of the LRV of the Apollo era, unmanned, without totally 'remote' control.

Doug