Northward autonomous rover challenge!

[18Watt]

@Pds314, here's an update.  It blew up again, as always while I was away from my computer.

Spoiler

Did not get insanely high altitude.  The highest speed of 175 m/s is odd.  (Edit:  the 175 m/s is normal I guess- it says that even when I haven't blown up..)

I didn't recognize this right away, but it starts with 'AutoRov3 crashed', not one of the parts.  I now suspect the probe core is falling off it's mount somehow.  Note that I'm on very flat terrain.

 

Ok, here's a drop test, with some interesting results.  I dropped it from 150 m over the runway:

Spoiler

Drop test 1, from 150 m. 

Many of the parts exploded, but I think the probe core actually survived.  You can see some debris on the runway, so some parts survived at least.

Drop test after making a few changes-  I applied Rigid Attachment to the probe core, and to some structural panels supporting it.  Landed on it's wheels, no damage at all.  During additional tests, it occasionally landed on the nose cones or other parts, which did blow up, but most of the parts survived the 150 m drop test.

I applied Rigid Attachment to the probe core, and the gold structural panels supporting it.  As you can see, the probe core is mounted to the servo via two cubic octagonal struts.  I think I'm going to eliminate the cubic octagonal struts- I've seen them generate bizarre physics effects, at least in previous versions of KSP.

If you have any other ideas, let me know.  I'll see if I get better results from attaching the probe core more securely, and eliminating the cubic octagonal struts.

Edited July 2, 2022 by 18Watt

[18Watt]

Making good progress.  My current run just passed 60N.  I haven't completely solved the random exploding issue, so I have my fingers crossed I'll make it the last 30 degrees.

I'm hitting most of my turns fairly accurately, although the errors are starting to add up.  And the errors get larger as you get farther north.  Using a control device mounted to a servo works fairly well below 30N, but the farther north you get, the less accurate it is in tracking.

Not sure, but I think one reason I'm hitting my turns fairly accurately is because I'm keeping a fairly constant average speed.  Which may be because of a serious accidental design error I made- I'm using 6 small rotors, but forgot to set half of them to DEPLOY.  So half my rotors are providing thrust, the other half are providing nearly equal drag.  ~20 m/s is about as fast as I can go (that's what I'm using..), but it seldom goes below 14 m/s, and only very rarely goes above 26 m/s.  I'll document that design flaw when I post the run results.  (Er, I mean I'll detail my genius design- definitely NOT a serious design flaw.  Nope..)

Edit:  Forgot to mention that I've hit TWO trees head-on, and had to start over.  In each case, only the nose-cone was destroyed, the rest of the rover probably could have continued, although the timing would be messed up.  But sadly, they became stuck, no way to free them without using manual controls.

 

 

Edited July 3, 2022 by 18Watt

[18Watt]

I'm marking my path by having a KAL bring up KerbNet every 1500 seconds.  That brings up the KerbNet dialogue, but I still need to actually click to drop the marker.  @Pds314 said he didn't have a problem with that, I'm not actually controlling the rover.

While staring at my screen for hours on end, I did come up with a possible way to mark my path without touching the keyboard.  I could drop probe cores periodically using decouplers or docking ports.  To drop while moving, each probe core would need to be encased in structural components.  I'm pretty sure at ~20 m/s I could encase probe cores well enough to drop them from the rover.

Ok, maybe not the most practical solution.  But I'm currently dropping a KerbNet marker every 1500 seconds.  I've dropped about 20 of them, so I've had to click to drop that many times.

[Pds314]

9 hours ago, 18Watt said:

Making good progress.  My current run just passed 60N.  I haven't completely solved the random exploding issue, so I have my fingers crossed I'll make it the last 30 degrees.

I'm hitting most of my turns fairly accurately, although the errors are starting to add up.  And the errors get larger as you get farther north.  Using a control device mounted to a servo works fairly well below 30N, but the farther north you get, the less accurate it is in tracking.

Not sure, but I think one reason I'm hitting my turns fairly accurately is because I'm keeping a fairly constant average speed.  Which may be because of a serious accidental design error I made- I'm using 6 small rotors, but forgot to set half of them to DEPLOY.  So half my rotors are providing thrust, the other half are providing nearly equal drag.  ~20 m/s is about as fast as I can go (that's what I'm using..), but it seldom goes below 14 m/s, and only very rarely goes above 26 m/s.  I'll document that design flaw when I post the run results.  (Er, I mean I'll detail my genius design- definitely NOT a serious design flaw.  Nope..)

Edit:  Forgot to mention that I've hit TWO trees head-on, and had to start over.  In each case, only the nose-cone was destroyed, the rest of the rover probably could have continued, although the timing would be messed up.  But sadly, they became stuck, no way to free them without using manual controls.

 

 

WHOA! 60 DEGREES NORTH?

[18Watt]

22 minutes ago, Pds314 said:

WHOA! 60 DEGREES NORTH?

Passing 87N right now.  No trees to hit this far north, looking good.

[18Watt]

89 degrees 59 minutes 14 seconds North!  2 days 2:45:00.

The route I took:

Spoiler

All legs and turns are based solely on time.

• Leg 1  000 (N)  for 3,120 s.  This crossed the bay to the north of KSC.
• Leg 2  058 deg. for 8,820 s.
• Leg 3  007 deg. for 4,080 s.
• Leg 4  053 deg. for 3,300 s.
• Leg 5  351 deg. for 8,160 s.
• Leg 6  353 deg. for 7,200 s. 
• Leg 7  335 deg. for 10,800 s.
• Leg 8  315 deg. for 1,440 s.
• Leg 9  000 (N) until reaching the pole, or close to it.

Journey photos 1

Spoiler

Starting out from KSC runway, pointed north.  Before I get going I need to:

• Turn on the lights.  (Not that they'll do me any good...)
• Bring up both KAL controllers.  One controls heading, the other cycles the bay doors and 'drops' KerbNet markers.
• Set the prop blades to 9 degrees (actually 9.35###).
• Engage SAS, and lock it to track Normal.
• Throttle up, and start both KAL controllers.  The heading one runs for about 50,000 s, the other one repeats every 1,500 s.

Heading north, crossing the bay.  My speed drops to 14 m/s.  Would have been better to drive around the bay, but I wanted to make sure the amphibious capabilities worked- because I thought I'd need to cross some rivers at some point.  Turns out I never really needed amphibious capabilities- there are no rivers to cross on the route I took.

At 1,400 s. on the right KAL ('Doors KAL') the doors open to 15 degrees for 20 seconds.  There's a lot of RTGs in the cargo bay.  After about 3 hours they start to overheat.  Periodically opening the cargo doors solved that problem nicely.

Next, at 1,430 s. on the 'Doors KAL', it brings up KerbNet.  The KAL won't actually place a KerbNet marker, I need to click the 'Waypoint' button every time.  But at least the KAL prompts me to drop a marker at regular intervals.

KerbNet markers starting to add up.  Perhaps I should have only dropped a marker at the points where I changed heading- that would also have been easy to implement with the KAL.  Maybe next time..

Cruising along, hoping my path keeps me clear of steep mountains.

Night.  The creepy trees start attacking me.  I escaped this one..

Just thought this was a neat shot of Mun.

Lots of markers now.  But I'm only at 23N, still have a long way to go.

Journey photos 2

Spoiler

I was really hoping to miss that peak in the background, and I did!  My course is turning out to be very close to what I had planned for.  Although I seem to be getting a little farther north ahead of schedule.  That could be a problem.

In the next photo, you can see that eventually I need to turn left- before I hit the water.  I'm ending up slightly farther north than I anticipated, and if I end up too far north I won't make the left turn soon enough.  Nothing to do about it but wait to see how far off course I end up.

I tried to avoid terrain any steeper than about 10 degrees.  The rover will handle slopes up to 20 degrees if needed, but the risk of tipping over does increase.

29 m/s is about the fastest the rover ever hit, thank goodness.  The rover started catching air over jumps above about 25 m/s, which is not good at all- keep in mind I can't correct my attitude if I end up airborne.  It lands how it lands.  So getting airborne is something I wanted to avoid at all costs.  The costs being the painfully slow average speed of about 20 m/s.

Continuing up the valley, getting closer to the left turn.

I did overshoot my turn by a degree or so, but had allowed just enough margin for error that my path still kept me clear of the ocean.  The lakes in the next photo are right next to the ocean, so it was very nearly a failure.  But I made it ok.

Trees still jumping out at me at night.

Another one!

Holy smokes!  I think I'm getting close!

Journey photos 3

Spoiler

Ok, I'm on the snowpack!  The scariest part of the journey is just ahead:  I need to travel down some steep terrain to get to the flat polar ice.  The rover is pretty stable, but during testing there was a 1/3 chance of rolling over during the descent.  So fingers crossed.

Can't really see the rover, but you can see the slopes I'll need to get down to get to the flat polar ice.

To complicate things, this is what my tracking system does at high latitudes.  It should be locked to NORTH, but it starts to deviate considerably on slopes.  It also has an increased tendency to roll over when not heading due north and also on a left or right slope.  So going down the hill is a nail-biting experience.  Oh, and since I'll be going downhill, my speed will shoot up to 25-30 m/s, also not good.

Well, I made it this time.  Phew.

Smooth sailing from here on out.  Just keep it pointed north.  No more trees, no more rough terrain.

Another shot of the cargo doors venting the cargo bay every 1,500 s.

Almost there.

Really close now!  The KAL drops one last KerbNet marker.  Actually, it would just keep dropping them every 1,500 s. forever.

The closest I saw was 59 d. 59' 14".  By the time I took this screenshot it had overshot that value.  But you get the idea, I definitely got pretty close to the pole..

Mistakes

Spoiler

There were a lot of mistakes made, and numerous runs which did not go the distance.  Hit a tree head-on, actually the second time I did that.

The collision destroyed the nose cone.  The rover would probably still function fairly well, but there's no way to make it un-stuck.  So I have to start over.  Oh well, I was only 4 hours into this run..

This is the next run after the tree crash (which eventually finished..).  And here's that EXACT SAME TREE!!!  I was surprised that my navigation system was accurate enough to nearly hit the same tree TWICE, after 4 hours of driving.  I missed the tree this time.

But I didn't miss it my much!

20 m/s was a good average speed.  Oddly, at one point during testing I realized that I couldn't get the rover going much faster than 20 m/s.  With 6 rotors, each having 8 blades, I should have been able to go much, much faster.  Well, turns out I forgot to deploy the 3 aft prop discs.  So 3 of the props are spinning flat, just producing drag!

Not sure if that helped to maintain a constant speed or not.  But I was indeed able to keep my speed between 15 and 30 m/s, even on moderate slopes.  I realized my mistake before making the last run, but at that point the rover was so predictable that I didn't want to change anything.  So I just left my mistake in place.

I'm playing 100% stock, with the MH + BG expansion packs, no mods at all.  With an average speed of 20 m/s, I suspect my time to the pole will be beaten sooner or later.

[QF9E]

5 hours ago, 18Watt said:

89 degrees 59 minutes 14 seconds North!  2 days 2:45:00.

Incredible! What an epic journey, I'm glad you made it to the pole!

My kOS craft isn't fairing so well, but I am intentionally not using any waypoints, as that would make the challenge trivial. I'm running a fairly simple program that just aims North but avoids obstacles: if below 50 meters or above 1000 meters it tries to go around. I'm also using an autothrottle to avoid getting launched by humps in the terrain but that one needs some fine-tuning as my latest attempt crashed at 16 degrees North after going over a hill too fast. My craft does have a self-righting mechanism but it tends to get damaged on hard landings.

My rover uses rover wheels, but seeing your attempt I think I'll change to landing gear and props, as landing gear is less prone to being damaged. It also makes a self-righting mechanism redundant: instead you can simply carry a second set of landing gear on top of the craft, and deploy those when coming to rest upside down. And you can swim across water, which comes in handy as well.

[18Watt]

I am working on an improved rover.  Currently in the testing stage, but looks promising.  It maintains a fairly constant 45 m/s, and can handle jumps somewhat confidently.  It still won't handle trees though..

I suspect I'll still need to keep to relatively flat terrain, but I'm looking at how it handles the mountains.

 

[18Watt]

On 7/4/2022 at 7:41 AM, QF9E said:

I think I'll change to landing gear and props, as landing gear is less prone to being damaged.

My personal opinion is it’s easier to maintain a steady speed with props.  At least I haven’t figured out how to maintain a steady speed with rover wheels.

[sevenperforce]

On 6/28/2022 at 6:34 PM, Pds314 said:

Well or player action/axis groups. Or other KALs, but there's not really sensor inputs.

No sensor inputs, but you can definitely do some fun stuff with chaining KALs. I once built a bunch of cluster bombs which launched fireworks on impact with the ground by using two KALs: one to command the fireworks to fire and the other to stop the first KAL. The cluster bombs deployed parachutes which lowered the impact speed to just above the impact resistance of the part the second KAL was mounted on.

I wonder if it would be possible to use old-fashioned stock hinges to couple disconnected vehicles so one vehicle would force the rover to maintain level movement, etc.

[mister_goo]

Jet-powered rover. Starts from 6 deg, out of fuel at 20 deg. Flipping is not a problem, and it can survive falls from cliffs (the terminal speed is less than the impact tolerence).

[18Watt]

Ok, I've noticed a problem:  It is not possible to go NORTH in stock KSP.  Yes, I thought it was possible by locking SAS to the Normal vector.  But that does NOT point North!

In surface mode, the Normal vector points perpendicular to the plane that the equator is on.  If you are on the equator, the Normal vector does indeed point roughly north.  However, the farther north you get, the greater the vertical component of the Normal vector becomes.

Here's some screenshots as examples.  I've Hacked Gravity to the minimum, to show where the ship points when locked to the Normal vector.

Spoiler

1 - Rover at KSC (near equator), zero gravity, Surface mode, locked to Normal.  Pointed North.

2 - Rover at KSC, Orbit mode, locked to Normal.  Also pointed North.

3 - Now the rover is at 45N, Orbit mode, locked to Normal.  Pointed North.

4 - Rover at 45N, Surface mode, locked to Normal.  Pointed North, but also +45 degrees above horizon.  This is why my rovers keep pitching up when airborne the further north I get.

5 - As another example, the next shot is at 60N, Orbit mode, locked to Normal.

6 - And here's 60N, Surface mode, locked to Normal.  The farther north you get, the worse the effect is.  At 89N, Normal means pointed 89 degrees above the horizon, and 1 degree towards north.

I'm scratching my head trying to come up with solutions in Stock KSP.  Here's some ideas I've had, and the problems I see with them:

• Don't use SAS.  Without SAS, the rover won't try to pitch up, with increasing severity at higher latitudes.  The probe core however is still trying to aim 'up' instead of 'north', so you still get sloppy navigation, especially on slopes.  This is why rovers constantly try to angle UP slopes, getting worse the farther north you are.
• Find a Mod.  Not sure if there's a mod out there which will actually point north- I suspect there is.  I don't use mods, but would possibly consider one if it solves this problem.
• Disable Pitch in Reaction Wheels.  I'd consider this modifying the game, although it would be pretty simple to alter the configuration file for the Reaction Wheel, to disable pitch torque.  As with the first solution mentioned, it still leaves the errors that come with the probe core trying to align with an increasing vertical component- turning up slopes, sloppy navigation.
• Add a Hinge.  I'm currently playing with this solution, using a hinge to angle the probe-core upwards the farther north I am.  This can be controlled by a KAL, so I'd still be within the rules of the challenge.  Somewhat reduces the tendency to angle UP on slopes, but does not solve the sloppy navigation at high latitudes.  In fact, it seems to be making the navigation even more sloppy.
• Go Slower!  Yeah, that actually works, kinda.  My first attempt did indeed make it to the pole, by going so slow that it never (or seldom) got airborne.  But 20 m/s is painfully slow.  I have a rover that can do this challenge at 55 m/s no problem, if I could only figure out how to make it point North.
• Target Track to a Flag.  (thanks @QF9E!)  The flag is planted on the North Pole.  Still have a 45 degree angle difference at the equator, but gets lower the farther north you are.  One minor disadvantage is that there are fewer probe cores that can track targets, the smallest is a 1.25 m format.  Currently testing, but this method looks promising.

My current rover can handle most terrain at 55 m/s no problem, probably even faster.  Unfortunately, it does get airborne frequently, and at high latitudes tends to pitch up and go flying much farther than what is safe.  I need a way to make it point NORTH, instead of aligning with a line perpendicular to the plane the equator is in.

Anyone have any further suggestions?

Edit:  No, using Orbit mode does not work well.  In orbit mode, 'Normal' means roughly 90 degrees away from whatever your current heading is.  So in Orbit mode the Normal vector only points North when you are stationary.  When you are moving, the Normal vector (in Orbit mode) no longer points North.  So that won't work.

Edited July 8, 2022 by 18Watt

[QF9E]

Could you perhaps do something with a flag planted at the North Pole and using that as target?

[18Watt]

5 hours ago, QF9E said:

Could you perhaps do something with a flag planted at the North Pole and using that as target?

Yes, and that may be the best solution for my rover.  There is still the problem of not pointing directly at the pole- in this case a Flag Pole, not the North Pole..  

One advantage would be the angle would be the worst at the equator (-45 degrees), and would progressively get better as you approached the pole.  Another advantage is it does not require mods, which I'm trying to avoid.

[18Watt]

@QF9E, I'm currently testing the method you suggested.  I'm near KSC right now, so the angle is the worst it will be during the trip to the pole, and the rover seems to be tracking fairly well!  I did add a hinge on top of the servo, to keep the probe core angled (somewhat) towards the target.

It's working better than using a hinge to track the Normal vector (pointing up more as you get farther north..). I'm hoping the tracking only gets better at higher latitudes- I'm guessing it will.

 

[18Watt]

Second autonomous polar run complete!  4 hours 53 minutes.

For this second entry, I went for a higher speed- 20 m/s takes way too long.  I built a rover capable of traveling at 50 m/s+, but ran into issues with attempting to track north.  With my 'slow' rover, jumps were limited.  However, at 50+ m/s jumps are frequent.  Tracking to the Normal vector points increasingly above the horizon, at 45N the rover wants to point 45 degrees above the horizon.  I detailed these issues in a previous post.

I ended up using a method suggested by @QF9E, planting a flag on the north pole, and using that as a target reference.  So, step #1 was planting a flag on the North Pole.

This rover is not amphibious, so I had to navigate around the bay north of KSC this time.  Yes, I’m sure this rover isn’t amphibious- I accidentally ‘navigated’ into the sea a few times.  It doesn’t sink to the bottom, but also doesn’t float high enough to keep the props dry. 

Spoiler

I flew a crew up to the North Pole to plant a flag.

Flag planted at 89.997N.  That's close enough for me.

Starting out:

Spoiler

I have two KAL controllers on board, although I only used one of them.  The heading servo keeps the rover pointed north.  The Elevation hinge angles the probe core up or down, to keep it pointed at the target, which reduces errors.  I have the probe core set to drop a KerbNet marker every time I make a heading change.

Sitting on the runway, ready to start.  I need to set the props to 12, throttle to 2/3, select the flag as a target, lock SAS to target tracking, release the brake, and start the KAL controller.  After that, I don't need to touch the controls again at all.

Leaving KSC on a 305 degree heading (-55).  The hinge is angling the probe core 35 degrees down from vertical.  The KAL will adjust that angle at scheduled time intervals during the journey.  Looking good so far.

At the first turn, the KAL drops a KerbNet marker, just like it was supposed to!

20 minutes into the journey, I've passed 5N and dropped 3 KerbNet markers at heading change locations.

More screenshots from the drive:

Spoiler

Catching some air.  The rover handles small jumps fairly well.  I had to add fins on the back to prevent it from fishtailing sometimes during landing.  I also set the rear wheels friction to 3.5, which also helped prevent fishtailing.

Whee!

Dropped another marker at a turn.

My nemesis- TREES!  They are by far the most dangerous part of the journey.

The orange marker is at the south end of what I call 'Tree Valley'.  Because it's full of trees.

Moon in background.

Tree!

Two more trees!  I was rarely out of sight of at least one tree in this valley.

I wanted to get a screenshot of the mountain in the background, and wasn't paying attention, went whizzing right past this tree.

The purple marker is the end of 'Tree Valley'.  There are still some trees on the next leg (I even hit one of them..), but not as bad as Tree Valley.

Final legs:

Spoiler

Even though I'm not in Tree Valley anymore, there are still trees.

This one was really close.

I can see the snowpack ahead.

The slope to the icepack is kinda rough and hilly.  This is where rovers that track to the Normal vector don't do well.  My slow rover had trouble even at 20 m/s.  This rover, tracking to a flag, had no issues even at 60 m/s.

The biggest jump of the voyage.  A little nerve-wracking, but it landed ok.

KAL dropping the last KerbNet marker.  From here on out, I'll be tracking due north.  No more turns, and no more trees.

Screenshot showing the final KerbNet marker.

89N!  Only 10.4 km to go!  (One degree on Kerbin is about 10.47 km..)

89.5N!!  Only 5.2 km to go!

89.9N.  I think we're going to make it!

There's the flag.  I wonder what will happen when I hit it.

Not sure if I missed it or just drove right over the top of it.  But as you can see the flag is still standing.  4 hours 53 minutes to reach the North Pole.

After passing over (or just near?) the flag, the rover just entered a circle around the flag, at a distance of 250-350 m.

A few extra screenshots:

Spoiler

Screenshot showing the markers dropped at each turn I made.

This is what happens when I hit trees, not good.  This impact took out a wheel and several propeller blades.  Rover is still limping along, but it won't be able to finish the journey.  This tree impact was at about 60N, so that was a lot of time wasted.

Rover navigation setup:

Spoiler

Here's the navigation schedule I used.  I figure my average speed was about 59 m/s.

• 0s (Start)  305 degrees for 180s
• 0 degrees (North) for 120s
• 30 degrees for 950s
• 56 degrees for 1,500s
• 8 degrees for 1,150s
• 30 degrees for 1,300s
• 353 degrees for 4,900s - Tree Valley
• 335 degrees for 4,500s
• 0 degrees (North) until reaching the pole

And here's the schedule I used for angling the Probe core downwards:  (Times are total from the start..)

• 0s (Start)  35 degrees down
• 5,200s  30 degrees down
• 7,200s  25 degrees down
• 8,300s  20 degrees down
• 10,100s  15 degrees down
• 12.100s  10 degrees down
• 14,100s  5 degrees down - Stayed at 5 down for the rest of the voyage

The KAL was set to drop KerbNet markers every time the rover changed course.

Trees were a big problem.  With a slower rover it would probably be possible to build a rover that could bounce off trees, but at 60 m/s that becomes more difficult, so I didn't even bother.  Instead, I just hoped for the best.  I hit several trees in the process.

Initially, I just started over without changing anything, hoping that random bounces would cause me to miss the tree the next time through,  That didn't work- on this entry I managed to hit the exact same tree twice!  So I had to change my strategy.

My new strategy was if I hit a tree, I would adjust the timing of the previous turn by a few seconds.  That normally worked, but on two runs I missed the first tree, only to hit a different tree, and have to start over after making a second adjustment to my turn timing.  Not a huge frustration when hitting trees early in the voyage, but I hit one at about 60N, which is about 3 hours into the drive.  That was really frustrating.

@Pds314, that's my latest entry..

Edited July 9, 2022 by 18Watt

[18Watt]

@QF9E, your suggestion of navigating to a flag planted on the North Pole worked quite well, thank you!

Here’s a suggestion if you are having trouble maintaining a constant speed-  Use a lot of propellers.

My latest rover would have been more than adequately powered using one propeller, or two to offset torque effects.  However, I had 8 propellers mounted on it- way more than actually needed.

The extra propellers very much help to maintain a constant speed, in my opinion.  I set the rotation to max speed (460 for the smallest rotors I think), and adjust speed using blade angle.  Max rotation speed is of course not the most efficient setting for propellers, but works well if you want to keep a constant speed, and don’t care about energy efficiency.

My (recent) rover was very lightweight and had 8 propellers.  I was able to maintain very close to target speed (~59 m/s).  Uphill I rarely went below 56 m/s, downhill I don’t think I ever went above 62 m/s.  I did avoid the steep mountains, but during testing even steep mountain slopes did not severely affect my speed.

Otherwise, perhaps MechJeb or another mod can aid speed control using rover wheels.  I don’t use mods, and wasn’t able to figure out how to keep a constant speed with stock rover wheels.