Robot Rumble 2.0 - Robot Combat Simulator - Under Development

[cjbruce]

The 28May2020 Bleeding Edge Build is up:

http://www.robot-rumble.com/bleedingedgebuilds/

Just a few tweaks from the build two days ago:

[Bug Fix] Pause menu pauses now.  [TODO] The rest of the menus still need to be consolidated between the different arenas.

[Bug Fix] Belt pulleys and chain sprockets were improperly scaled in some cases.

[Bug Fix] Scaled down the moment of inertia to 30%.  This should make Poison Arrow drivable.

[Bug Fix] Fixed scale of the high speed spinner collider.  It was set to 5% instead of 95% of the spinner's maximum radius.

[Bug Fix] "Reset Bot" in the test cage now resets the telemetry as well.

[Changed] Changed the description for the A28-400 to say "middleweight".

[Somebody]

Hmm.  I hadn’t thought to put a controllable current limiter into the game.  That would probably be helpful.  We don’t do it with our own ESCs, and I didn’t realize it was a thing that people do for brushed ESCs.

Right now brushed ESCs are completely unlimited.  I ended up artificially reducing the torque for drive motors to dull the initial current (and torque) spike.  Drive motors were so torquey initially that the simulation couldn’t keep up and it was causing the lighter robots to jump unrealistically around the arena.

It wouldn’t be difficult to add a current limiter to ESCs.  60-100 amps is low!  I haven’t built a drive system with the A28-400s yet, but I suspect most robots will exceed that unless they are geared down a lot.

It sounds like the brushless motor model might be reasonably close to real life.  I limited current to some suitably small value.  I can’t remember the exact number, but it is something like 25% of what the stall torque would be for a brushed motor.  You might try the brushless motors in game.  They have an insane top speed but their current limiting means they have to be geared way down.  If we can find the numbers for the motors you are using I would be happy to put them in the game for testing purposes.

We began to swear by current limiting when we were one of very few teams who could keep 48V Ampflow motors alive without them burning up. Their stall current was 700A or something crazy so just the instantaneous spikes were enough to detonate them. Ours ate themselves later due to high-speed issues, but even then the current-limiting prevented any fire on what likely would have otherwise been a short-circuited situation. Twice. The Ragebridge2 controllers we used had it as well as the Kelly controllers, and the VESC6 brushless controllers that we'll be using for 2020 also have current limiting. At this point I see it as a requirement to make a reliable robot.

As for the 60-100A, that was due to compounding faults. The first was under-gearing the drive, which was an intentional choice to save weight. The main issue was that when calculating weapon spinup, we didn't calculate the effects of motor efficiency. Since the Ampflow motors were outputting 30% of their inputted energy as heat, they were swallowing 30% more battery power during a fight than we expected. So we had to re-wire the robot from 2 batteries for drive/2 batteries for weapon to 1 battery for drive/3 batteries for weapon. We did this after the Jasper fight. This meant that 60A was the max current we could run to guarantee that our drive batteries wouldn't die. After fighting Gigabyte and Bronco for three minutes each, we felt comfortable enough with our in-fight battery usage to bump it up to 80A for HyperShock and 100A for Whiplash, if my memory serves correctly. I agree it was super low, and it's why we've done a lot of work to swap away from those parts for 2020. But the rules don't say you have to be fast, you've just got to move!

Here's a link to the motors we're using this year:

https://maytech.cn/products/brushless-hall-sensor-motor-mto8085-160-ha-c

I think limiting the torque of brushless motors, especially on startup, is the right move. The real answer to brushless motor torque is probably some kind of curve that you would need to dyno-test motors to develop, and even then is dependent on ESC choice, ESC settings, the gearing style in the robot, and probably a hundred more things that we don't even know yet. It's telling that even the teams don't truly know how they will perform until it's tested in the robot. Otherwise, sorry to add another wrinkle of complication to your simulation efforts. You're doing awesome work!

[min440303]

Was still playing May 26th and found something interesting.
 

Default Windows desktop 64-bit 2020-05-29 00-49-14-122.mp4

 

[cjbruce]

No worries!  You definitely have given me a few more things to chew on.  There is no thermal or electrical model for the battery or ESC.  I need to decide how far to go with this before it isn’t fun.

The Maytech motors are so tiny!  Wow!

[cjbruce]

Was still playing May 26th and found something interesting.
  [ Quoting of attachment images from other messages is not allowed ]  

This is an interesting case.  I had assumed that anything that doesn't have limits would be a spinner and would get a high speed spinner collider.

I need to think about how to handle this case.  Maybe increase the speed at which the high speed collider becomes active...

[min440303]

Poison Arrow did get more drivable in the new build which was great, and I found hits were getting much smaller, it was almost like no impact there, was that normal because of some tweaks?

[kix]

Some feedback i suppose:

The gyro has improved by a bit

The hits are kinda still underwhelming and i still feel like the hits are delayed/not happening, the horizontal shell i made dishes out no damage, idk if its the speed or anything, but i guess i can drop the botfile.

The self righting with the disc is not working at all, when it hits the floor, it just slows down to a halt, ant then the motor overheats:


Also the game is more unstable, crashes are more often

[min440303]

The hits are kinda still underwhelming and i still feel like the hits are delayed/not happening,

The hits were great in the last two builds I think. They are just massively weakened in this new one.

[cjbruce]

Hits are coming soon!

This morning I scratched out a strategy/math for computing the required amount of bite at a given speed.  Shower = less bite required for a good hit, also more bite

As the weapon spins up, a deeper bite is required, but less bite will be available.

Beyond a certain critical RPM and forward speed it will be impossible to bite and the weapon will not do any damage.

[cjbruce]

Crashes?  Uhoh!  Where are crashes occurring?  Can you replicate them reliably?

[kix]

Crashes?  Uhoh!  Where are crashes occurring?  Can you replicate them reliably?

Crashes occur mid match. How to replicate? Hit a bot. Idk if its the Robogames arena or the game, use the 2 bots i sent. If you want the arena, i can PM you the arena, as its not really planned to be released as of now so i wont post it here

[cjbruce]

New bleeding edge build is out (29May2020):

http://www.robot-rumble.com/bleedingedgebuilds/

Updates:

[Added] Added Ultimate Tensile Strength (UTS) to ArmourMaterial. This is a material property that can be found online for any given material. It will be used together with the spinner's current KE to determine the amount of bite required for a "good hit".  The idea is that harder armors require less bite.  You would have to bite really deeply into UHMW in order to get a good hit.

[Changed] Revised the motor thermal model to make overheating less likely. All motors in the game are now assumed to use NEMA Class F silicone insulation in their windings, rated to 155 degrees.  This is based on testing with my Tombclone and the ME0708. The electrical system now includes both temperature-dependent and temperature-independent resistance. This paves the way for a more complete electrical model of the battery, ESC, and connecting wiring.

[Temporarily Removed] Good Hits - Right now all hits in SpinnerCollisionHandler.cs are hard-coded to be glancing blows. Glancing blows need to be calibrated first. Then good hits. Then code needs to be implemented to distinguish between the two. NOTE - The mathematical framework for determining the difference between a glancing blow and a good hit is already in place. Hopefully it will be ready to go in the next commit or so.

[Bug Fix] Angular momentum and moment of inertia were incorrectly reported after the MOI adjustment implemented in the most recent build. This drastically reduced the amount of impulse applied to the two objects when a spinner hit.  Min and kix, I believe this was the cause of the "weakened hits".  They should be restored to more than their previous amount of oomph.  I will be adjusting them again now that there is a plan for spinner bite.

[Bug Fix] Updated telemetry for spinners to eliminate instances of infinity and NaN. Now things should correctly display with their actual values and update to zero when the motor is destroyed.

[Tweaked] Increased speed at which high speed spinner collider becomes active to 300 RPM (5 revolutions per second). Hopefully this eliminates the problem min was seeing with his robot.

[Bug Fix] Ready button not appearing in Test Arena

[Bug Fix] Main Menu button not working in Test Arena

[Bug Fix] Slider Values in setting menu not updating

[Bug Fix] Settings not saving in Main Menu

[CHANGE] Updated Workshop to use new settings menu

[CHANGE] Slider Values now display whole numbers only

[Bug Fix] Aspect Ratio mismatch to Resoution

[Bug Fix] Resolution being downscaled on scene transition

[CHANGE] Only allow 16:9 Resolutions (for now)

[kix]

[Added] Added Ultimate Tensile Strength (UTS) to ArmourMaterial. This is a material property that can be found online for any given material. It will be used together with the spinner's current KE to determine the amount of bite required for a "good hit".  The idea is that harder armors require less bite.  You would have to bite really deeply into UHMW in order to get a good hit.

I did notice a difference when a bot used UHMW to protect its front compared to Steel, so it works

[Changed] Revised the motor thermal model to make overheating less likely. All motors in the game are now assumed to use NEMA Class F silicone insulation in their windings, rated to 155 degrees.  This is based on testing with my Tombclone and the ME0708. The electrical system now includes both temperature-dependent and temperature-independent resistance. This paves the way for a more complete electrical model of the battery, ESC, and connecting wiring.

It is noticeable, the motors can be geared a bit higher now, which is good

[Bug Fix] Angular momentum and moment of inertia were incorrectly reported after the MOI adjustment implemented in the most recent build. This drastically reduced the amount of impulse applied to the two objects when a spinner hit.  Min and kix, I believe this was the cause of the "weakened hits".  They should be restored to more than their previous amount of oomph.  I will be adjusting them again now that there is a plan for spinner bite.

Again, fairly noticeable as vertical spinners are now able to self Right

[Bug Fix] Resolution being downscaled on scene transition

This is still not working sadly, only place it works is in the multiplayer battle bot selection scene.

[min440303]

The hits are even better than before which is brilliant.

And it's always me coming up with some weird stuff in the bunch of new builds.

Idk if it's only me that's having the problem. For 2WD bots (not only verts, all bots with only two wheels in the back), all of them are fast as hell in the new build, and I can't gear the motors to slow them down because of the overheating. if I use batteries with lower volts, the speed is lower, but the driving is a bit awkward, and the bot keeps lifting its front which means the bot should be faster but something's stuck on the floor to prevent it from moving faster like the driving issues I've mentioned before.

Then it comes the situation like: The best gearing I've made is like the first vid below, Aftershock drives really fine when the temp rises up to 70~80, but before that the bot just overdrives madly.

High Volts with low gearing


Low volts with high gearing

[cjbruce]

I just tried Aftershock.  The 10S you are using is pretty high for drive.  According to the Robot Wars wiki Aftershock uses 6S for drive.  I wouldn't be surprised if you had overheating problems.

I swapped out the 10S for 6S, and it drives okay.  I tried messing with the gear ratio a bit, and I'm mostly happy with it.

I'm also really curious to see how it drives using analog sticks on a game controller rather than a keyboard.

[min440303]

I just tried Aftershock.  The 10S you are using is pretty high for drive.  According to the Robot Wars wiki Aftershock uses 6S for drive.  I wouldn't be surprised if you had overheating problems.

I swapped out the 10S for 6S, and it drives okay.  I tried messing with the gear ratio a bit, and I'm mostly happy with it.

I'm also really curious to see how it drives using analog sticks on a game controller rather than a keyboard.

How much did you gear it to?

[cjbruce]

Somewhere in the range of 8:1 to 12:1.

I think I'm okay with slower driving.  I'm old. :)

Also, I didn't mention it before, but temperatures will be higher in this build.  I increased the room temperature to 40 degrees. I simultaneously increased the temperature limit of the windings to 155.  You should have a lot more room before your windings overheat.

[min440303]

Somewhere in the range of 8:1 to 12:1.

I think I'm okay with slower driving.  I'm old. :)

Also, I didn't mention it before, but temperatures will be higher in this build.  I increased the room temperature to 40 degrees. I simultaneously increased the temperature limit of the windings to 155.  You should have a lot more room before your windings overheat.

Ya it should be. As spinner bites became a thing, everytime a spinner hits the floor the temp gets much higher. Especially for verts, sometimes a vert has to hit the floor over 5 times to self-right, although those are small bites, the temp goes up pretty fast. Not sure if it's a thing but I'll just post it here.

 

Default Windows desktop 64-bit 2020-05-30 19-59-12-729.mp4

[kix]

I just tried Aftershock.  The 10S you are using is pretty high for drive.  According to the Robot Wars wiki Aftershock uses 6S for drive.  I wouldn't be surprised if you had overheating problems.

I swapped out the 10S for 6S, and it drives okay.  I tried messing with the gear ratio a bit, and I'm mostly happy with it.

I'm also really curious to see how it drives using analog sticks on a game controller rather than a keyboard.

I should try using a controller for the bots tbh, i could give you feedback i suppose

[cjbruce]

kix, I'm testing now with Green Banana and Bloodspill.  It is a really good combination of robots and has allowed me to focus on how two robots should behave when they hit each other. 

Some numbers:

A "satisfying" collision impulse for these two robots is somewhere between 50-80 Newton-seconds.  A weak impulse is around 10 Newton-seconds.  Above 120 Newton-seconds and the two robots go careening around the arena uncontrollably/unrealistically.

Our sweet spot for satisfying hits is around 60 Newton-seconds.

The problem was how to visually distinguish between "good hits" and "glancing blows".  If they both look the same, players won't know when they have really connected vs just scraping up some armor ineffectually.  I think I've hit upon a difference:

Good hit = robots get pushed away in the direction of the spin

Glancing blow = robots get pushed away from each other radially

All hits can now involve around 60 Newton-seconds of impulse, enough to be satisfying but not enough to break the game.  Glancing blows will be better for self-righting a vertical spinner.  Good hits will cause damage.