Modify RR vs Pedro into a table format #46

BaronClaps · 2025-07-07T21:58:31Z

Cleaned up some of the pros/cons so they’re worded less absolutely, especially for Pedro.

Being a pro for one isn’t an automatic con for the other.

Fixed the AdvantageScope and path optimization sections to match what’s true right now.

BeepBot99 · 2025-07-08T01:16:28Z

Here are my thoughts.

Being a pro for one isn’t an automatic con for the other.

First, I want to say that I completely agree with this. It's not only a hassle to duplicate every point (and doesn't add any value or new information for the reader), but can also lead to things that are misleading. For example, (as explained later) AdvantageScope support may be a pro of Roadrunner, but it is not a con of Pedro.

Driving Faster

Wording less absolutely is good. This does the opposite of that. For example, changing from "can make your bot drive faster" to "makes your bot drive faster" is definitely not less absolutely. While I don't think saying "makes your bot drive faster" is good, I also don't like "can make your bot drive faster." This is because many good things about Pedro (and bad things about Roadrunner) are less definite, like "can make" and "potentially worse," where it's not true for the other side.

I think the best option is to make that one "easier to make your bot drive faster." Although it is probably true that both can go that fast (although I'm not sure about that), it is likely much easier in Pedro.

Less People

I think less people is a valid con and should not be removed. It is definitely something that could matter to many people, especially beginners. However, it is also true that while Pedro may have less, it is still very many. I think keeping it as "less people use it compared to Roadrunner" is completely fine, but possibly instead something like "less people use it compared to Roadrunner (although still a lot!)"

Mostly Manual

While it is true that Pedro tuning is more manual that Roadrunner's, I think "mostly manual" is false. It is unfair to count every single constant. A PID counts as one thing. Additionally, "32 manually tuned constants" is including ones that next to no one tunes. Even the Pedro developers. I think if you're going to decide how many things there are to tune you should have tuned Pedro before. I would count it like this:

Automatic
- Forward Velocity
- Lateral Velocity
- Forward ZPA
- Lateral ZPA
- Forward Localizer
- Lateral Localizer
- Turn Localizer (if using three wheel or OTOS)
Manual
- Translational PID
- Heading PID
- Drive PID
- Centripetal

While it is true that the manual tuners take longer than the automatic tuners, I would say time spent is roughly equal.

Time Consistency

I don't see why this matters. If it actually does and it's not just a buzzword, perhaps an explanation would help. Additionally, "speed is prioritized over consistency" is phrased like something bad, but going fast is good!

Optimizing Curves

Roadrunner does not optimize the curves.

Logs

The phrase "much more difficult debugging" cannot possibly be true. 99% of Roadrunner users do not use the logs for debugging. Additionally, it is extremely easy to log the pose yourself, if needed. I don't think this should be a con of Pedro, although being a pro of Roadrunner is completely fine.

AdvantageScope

"Full AdvantageScope support" is just a lie. The pull request has not been merged. Furthermore, even once it is merged, no AdvantageScope support is not a con of Pedro, just a pro of Roadrunner. I think removing it from Roadrunner pros is the right choice (as has been done in this PR) until the AdvantageScope PR is merged. It also looks like it was accidentally not removed from Pedro's cons.

Bugs

"Minimal bugs if any" is not true. Roadrunner has had its fair share of bugs recently. "Potentially buggier" was removed from Pedro's cons; I think this should be removed from Roadrunner's pros as well instead of adding "recently this has not been true." Neither is significantly buggy, and occasional bugs are expected.

Bulk Reads

This is not a con of Roadrunner 💀.

Correction

I think saying "potentially worse correction" is not true. Roadrunner has worse correction while following a trajectory.

BeepBot99 · 2025-07-08T01:38:03Z

I think something like this might be better, and less susceptible to misleading.

Feature	Pedro Pathing	Road Runner
Path Following	Reactive vector follower with 3 PIDs	Trajectory follower using motion profiling
Tuning	Half manual and half automatic	Nearly all automatic
Motor Write Caching	Used by default	Able to be implemented manually
Centripetal Force Correction	Built-in	None
Path Generation	Manually using control points	Automatic spline optimization
Quickstart	Nonstandard package	Standard TeamCode package
Bulk Reads	Able to be implemented manually	Used by default
Logs	Not built in	Automatic logging of every run
Priority	Speed	Time consistency

Notice:

I say things like "able to be implemented manually" instead of "must be implemented manually" for both sides. This makes it seem less like a dealbreaker and more like an inconvenience. Using "must" makes it sound like it's required to have a good experience.
No information is duplicated.

0verkil · 2025-07-08T02:59:28Z

Some comments:

Additionally, "speed is prioritized over consistency" is phrased like something bad, but going fast is good!

True. Also, speed means nothing if your auto is inconsistent.

Path Following: Trajectory follower using motion profiling

This doesn't make sense. Trajectory and path following are two different things. Roadrunner has access to both Trajectory (the more commonly used variant) and Path (the lesser known but still existent and performant feature) following (along with a host of other capabilities and extensibility), whereas Pedro is exclusively a path follower. Putting them both as a path follower fundamentally misrepresents the capabilities of the two libraries.

Centripetal Force Correction: Built-in / None

This also doesn't make sense. Centripetal force correction only matters if you aren't motion planning, since it's essentially compensating for the path in front of you. Does roadrunner's path follower have centripetal force correction? Honestly, not sure, but it doesn't matter much. Roadrunner's trajectory follower definitely does, and that's the one people use. What do I mean by this? Well, Roadrunner uses motion planning to ensure the robot will (theoretically) track the path with equivalent accuracy to arc odometry (which i believe is convergent proportional to n^2). This compensates for curves by planning for the robot to travel around curves! If done correctly (can't comment on this as I haven't looked), this is actually a substantially better solution than just extrapolating off the current curvature.

Path Generation

The word "automatic" for RR here is a bit weird. The paths are still manually tuned, with automatic generation. Yes, it is more "automatic" than pedro's paths, but not really in a meaningful way. Users still manually tweak their paths in the same way you would tweak the control points in a pedro path.

Suggested changes (something like this):

Following: Reactive vector follower with 3 PIDs / Suite of followers including a trajectory follower using motion profiling and path followers (this list probably should be modified by someone who uses rr more)

Centripetal Force Correction: [This section is removed in my revision, because it's not a meaningful point of comparison.]

Path Generation: [This section is removed in my revision, because it's not a meaningful point of comparison.]

0verkil

Feels like this needs more eyes. From my perspective this PR has a lot of biases, but I am not impartial.

0verkil · 2025-07-08T03:03:47Z

src/misc/pedro_vs_roadrunner.md


 **Cons of Pedro:**
- Newer, so potentially buggier.


Still true! (not as much as when this was first written, but that doesn't warrant taking it away).

I think the rewrite makes this much more the case

The rewrite isn't public so it wouldn't make sense to discredit based on an unreleased version

0verkil · 2025-07-08T03:04:24Z

src/misc/pedro_vs_roadrunner.md

- Newer, so potentially buggier.
- Less people use it compared to Road Runner.
- Tuning is mostly manual. <!-- 4 automatically tuned constants, minimum of 32 manually tuned constants -->
- Not necessarily time-consistent; speed is prioritized over consistency.


Confused why the line about time consistency was removed. This is necessarily true.

0verkil · 2025-07-08T03:07:35Z

src/misc/pedro_vs_roadrunner.md

 - Time consistent by default.
- Tuning is almost fully automated, making it difficult to screw up. <!-- 5 automatically tuned constants, 3 manually tuned constants -->
+<!--- Tuning is almost fully automated, making it difficult to screw up. 5 automatically tuned constants, 3 manually tuned constants ... PEOPLE still do? -->


Change seems reasonable, given influx of RR help requests. Comment is unnecessarily inflammatory.

Change seems reasonable, given influx of RR help requests. Comment is unnecessarily inflammatory.

++. Please refrain from making such comments in cookbook. There have been 2 such comments in this PR.

Remove only the , difficult to screw up part.

0verkil · 2025-07-08T03:08:27Z

src/misc/pedro_vs_roadrunner.md

 - Uses the FIRST-recommended standard coordinate system consistently.
- Works with SlothLoad immediately.
- Bulk reads by default.
+<!-- - Works with SlothLoad immediately. || this is a quickstart thing, not a library thing itself? -->


Pedro was exclusively a quickstart before it was a library. The argument for this change seems moot.

atp both quickstarts both have the org.firstinspires.ftc.teamcode so 🤷🏻

Agreed, although not for the same reason, unless I'm understanding it wrong. I don't think it being a quickstart before a library matters. However, the quickstart is still very much part of Pedro Pathing. Until Pedro 1.1.0 releases (where this is fixed), it should be on there.

Happy to remove the sloth load line, next time please reference the change to the quickstart in the PR, especially as the change was made after the commits.

src/misc/pedro_vs_roadrunner.md

0verkil · 2025-07-08T03:09:51Z

src/misc/pedro_vs_roadrunner.md

@@ -70,7 +64,7 @@ Guidelines for editing this page:
 Differences must be objective issues from a neutral point of view, not one sided.
 Ideally, people with biases in both directions should agree about these differences.

-Each difference will be listed twice, as a pro of one library and as a con of another. 
+// Just because something is a pro of smth doesnt mean its a con of the other...


Changes seem ok. Comment is unnecessary.

Also... why // 😭

It's already inside an HTML comment!

I think the idea that the comment promotes is fine, we should be clear about the idea that going forward, we should be careful about a pro of one library not being a con of another, but we should remove the ellipses at least, and it would be better to have this formally worded as part of an article prelude, as this sentence is very informal.

0verkil · 2025-07-08T03:11:45Z

Oh also - would like to point out again that this

Reactive vector follower with 3 PIDs

is meaningless buzzwords. What is a "reactive vector follower"? Call it a path follower T_T

BeepBot99 · 2025-07-08T03:21:11Z

Roadrunner has access to both Trajectory (the more commonly used variant) and Path (the lesser known but still existent and performant feature) following

This is not true in RR 1.0. It does not have a path follower. I think it should stay how I wrote it (except change Path Following to just Following or possibly Following Strategy).

The word "automatic" for RR here is a bit weird. The paths are still manually tuned, with automatic generation. Yes, it is more "automatic" than pedro's paths, but not really in a meaningful way. Users still manually tweak their paths in the same way you would tweak the control points in a pedro path.

Yeah, I agree, I just left it how it was originally. It also does not determine the shape of the curve automatically in any way; that is completely manual. Maybe just do "cubic Bezier curve" and "quintic Hermite spline."

BeepBot99 · 2025-07-08T03:21:34Z

is meaningless buzzwords. What is a "reactive vector follower"? Call it a path follower T_T

I agree with this too.

Froze-N-Milk · 2025-07-08T03:27:36Z

Can you please squash your three commits into one, or split them with different names if they are purposefully separate.

Force push to the branch and update the PR.

BaronClaps · 2025-07-08T04:15:27Z

Can you please squash your three commits into one, or split them with different names if they are purposefully separate.

Force push to the branch and update the PR.

@Froze-N-Milk Done.

BaronClaps · 2025-07-08T04:15:45Z

Oops, clicked the wrong button. 💀

dr-hextanium · 2025-07-15T02:05:10Z

src/misc/pedro_vs_roadrunner.md

+| **Path Following** | Path Follower | Trajectory follower using motion profiling | **Visualizer** | Web-based using a nonstandard (0 - 144) coordinate scheme | Code-based using the standard FTC coordinate scheme |
+| **Tuning** | Half manual and half automatic | Nearly all automatic |
+| **Motor Write Caching** | Used by default | Able to be implemented manually |
+| **Paths** | Cubic Bezier curves | Quintic Hermite splines |


i don't think this section is relevant

I get 5, 6, 7 but I think 8 should probably stay

Hex commented on 8 lol.
I think that it should stay because even though it's unlikely to influence someone's decision, it's still good to know and it can't hurt.

Hex commented on 8 lol.
Oh.

i think just listing the type of curves used offers no value without explaining their advantages/disadvantages

dr-hextanium · 2025-07-15T02:12:02Z

src/misc/pedro_vs_roadrunner.md

-**Pros of Road Runner:**
- Stable, minimal bugs if any.
- Time consistent by default.
- Tuning is almost fully automated, making it difficult to screw up. <!-- 5 automatically tuned constants, 3 manually tuned constants -->


don't agree

With what? The content being removed or the fact that it's being removed?

With what? The content being removed or the fact that it's being removed?

with what is being said I presume

j5155 · 2025-07-22T16:30:13Z

I think the table format is a good idea and generally gets rid of a lot of duplication. However, it seems like this PR has also removed a lot (or maybe all??) of the explanations for why the reader should care about these differences? I think it's important to keep those.

I've made a release/readme for advantagescope RR support and it's usable today: http://github.com/j5155/advantagescope
(upstream will hopefully have it in August)

Will add some detailed comments as well.

j5155 · 2025-07-22T16:33:49Z

src/misc/pedro_vs_roadrunner.md

-to dynamically converge to the target position. 
+| Feature | Pedro Pathing | Road Runner |
+|--------|--------|--------|
+| **Path Following** | Path Follower | Trajectory follower using motion profiling | **Visualizer** | Web-based using a nonstandard (0 - 144) coordinate scheme | Code-based using the standard FTC coordinate scheme |


What this following difference means needs to still be explained. Something like "Path Follower; faster by default and always tries to get back to the path" and "Trajectory Follower with Motion Profiling; consistently timed path by default and tries to catch up when it gets behind".
The visualizer difference looks good, though the line break seems to be missing?

j5155 · 2025-07-22T16:35:29Z

src/misc/pedro_vs_roadrunner.md

+| **Path Following** | Path Follower | Trajectory follower using motion profiling | **Visualizer** | Web-based using a nonstandard (0 - 144) coordinate scheme | Code-based using the standard FTC coordinate scheme |
+| **Tuning** | Half manual and half automatic | Nearly all automatic |
+| **Motor Write Caching** | Used by default | Able to be implemented manually |
+| **Paths** | Cubic Bezier curves | Quintic Hermite splines |


Put back the original wording, or something close to the original wording here. The difference is important: Pedro paths can be modified arbitrarily using control points, but the control points have to be set manually. Roadrunner paths have automatically optimized curvature based on the tangents of each path, but manually changing the curve of the path is a little harder.

j5155 · 2025-07-22T16:36:33Z

src/misc/pedro_vs_roadrunner.md

+| **Motor Write Caching** | Used by default | Able to be implemented manually |
+| **Paths** | Cubic Bezier curves | Quintic Hermite splines |
+| **Bulk Reads** | Able to be implemented manually | Used by default |
+| **Logs** | Not built in | Automatic logging of every run |


I think Replay is a better term then Log here, and I would still like the "readable with AdvantageScope" link that goes to https://github.com/j5155/AdvantageScope now that I actually have a release.

(update: the PR is merged now and support is official)

Froze-N-Milk · 2025-09-01T01:53:10Z

@BaronClaps could you please make some of the requested changes? I'd love to get this article up to date now that 2.0 is out.

0verkil reviewed Jul 8, 2025

View reviewed changes

BaronClaps force-pushed the patch-4 branch from 5fc6b5f to ee33b45 Compare July 8, 2025 03:57

BaronClaps changed the title ~~Pedro vs RR fixes to be more neutral~~ Modify RR vs Pedro into a table format Jul 8, 2025

Put information into tables

9ca8bd9

BaronClaps force-pushed the patch-4 branch from ee33b45 to 9ca8bd9 Compare July 8, 2025 04:06

BaronClaps closed this Jul 8, 2025

BaronClaps reopened this Jul 8, 2025

dr-hextanium reviewed Jul 15, 2025

View reviewed changes

j5155 self-requested a review July 15, 2025 05:04

j5155 requested changes Jul 22, 2025

View reviewed changes

Update pedro_vs_roadrunner.md

2d618d8

Modify RR vs Pedro into a table format #46

Are you sure you want to change the base?

Modify RR vs Pedro into a table format #46

Uh oh!

Conversation

BaronClaps commented Jul 7, 2025

Uh oh!

BeepBot99 commented Jul 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Driving Faster

Less People

Mostly Manual

Time Consistency

Optimizing Curves

Logs

AdvantageScope

Bugs

Bulk Reads

Correction

Uh oh!

BeepBot99 commented Jul 8, 2025

Uh oh!

0verkil commented Jul 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

0verkil left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

0verkil Jul 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Froze-N-Milk Jul 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

0verkil commented Jul 8, 2025

Uh oh!

BeepBot99 commented Jul 8, 2025

Uh oh!

BeepBot99 commented Jul 8, 2025

Uh oh!

Froze-N-Milk commented Jul 8, 2025

Uh oh!

BaronClaps commented Jul 8, 2025

Uh oh!

BaronClaps commented Jul 8, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

BeepBot99 commented Jul 8, 2025 •

edited

Loading

0verkil commented Jul 8, 2025 •

edited

Loading

0verkil Jul 8, 2025 •

edited

Loading

Froze-N-Milk Jul 8, 2025 •

edited

Loading