Improve EnemyBallPlacementPlay and convert it to use FSMs #3146

raymond212 · 2024-02-24T02:49:02Z

Description

Fixed a bug in sensor fusion where the ball placement position is not sent to non-placing team, resulting in the stadium obstacle in Stadium Geometry Obstacle #3121 not working.
Rewrote EnemyBallPlacementPlay to use the FSM architecture.
Re-designed the play as: if the ball is far from the placement point, then focus on avoiding ball placement interference by assigning AvoidInterferenceTactics (clones of MoveTactic) to move robots out of the way. AvoidInterferenceTactics override motion constraints, because sometimes the stadium motion constraint can actually trap robots and prevent them from moving to somewhere that doesn't interfere.

Once the ball is nearly placed, then we switch to a defensive formation with 1 goalie, 2 crease defenders, and 3 robots to stay near the ball without interfering (essentially our StopPlay behaviour). By doing this, we can transition into our other plays quickly once the game has been resumed.

The edge case where the stadium goes through the defense area is also handled well.

In the first and third gifs, there is a sensor fusion bug where the ball position does not get updated, possibly because the yellow placement robot is blocking the ball. This could be something to address in a separate issue.

Added validation to check that we do not have robots stay in the ball placement stadium for over two seconds.

Works well generally. Three robots stay near the ball without interfering, so that we can quickly transition into the play that follows, while we also have defensive robots as the goalie and crease defenders.

However, there are issues with some edge cases. Currently, I try to position the three robots on the other side of a line connecting the ball to the ball placement position. This could end up outside the field, which is problematic. To fix this, I think I will try to directly send these robots near the final ball placement point, while making sure to stay out of the stadium obstacle.

Also, the stadium obstacle is not fully effective at keeping out the goalie and crease defenders, in the edge case that the stadium obstacle goes right through our defense zone. While unlikely, this is technically possible and allowed under robocup rules 5.2. The best solution for this situation would probably be to use move tactics to force our defensive robots out of the stadium.

Testing Done

Tested in Thunderscope and enemy_ball_placement_play_test.py.

Resolved Issues

resolves #3018, resolves #2783, resolves #2870, resolves #2599

Length Justification and Key Files to Review

enemy_ball_placement_play_fsm.h and enemy_ball_placement_play_fsm.cpp for FSM implementation
robot_enters_placement_region.py for validation

Review Checklist

It is the reviewers responsibility to also make sure every item here has been covered

Function & Class comments: All function definitions (usually in the .h file) should have a javadoc style comment at the start of them. For examples, see the functions defined in thunderbots/software/geom. Similarly, all classes should have an associated Javadoc comment explaining the purpose of the class.
Remove all commented out code
Remove extra print statements: for example, those just used for testing
Resolve all TODO's: All TODO (or similar) statements should either be completed or associated with a github issue

…into raymond/enemy_free_kick

…placement

nimazareian · 2024-02-24T03:38:32Z

Good work on finding this bug!

With regards to this

Also, the stadium obstacle is not fully effective at keeping out the goalie and crease defenders, in the edge case that the stadium obstacle goes right through our defense zone. While unlikely, this is technically possible and allowed under robocup rules 5.2. The best solution for this situation would probably be to use move tactics to force our defensive robots out of the stadium.

I have a work in progress branch with a couple of core changes to the trajectory planner, including integrating #3058 into the planner. This should ™️ move the robots out of all non-robot obstacles, so hopefully all robots will move out of the stadium. You may have already done this, but it might be helpful to have a sim test with this edge case that I could run on my branch later as well.

If the ball moves quick, causing the stadium to move quickly, we might still receive a foul... but Im not sure if that is really avoidable.

nimazareian

I know this PR is still a draft, but quickly looked over it to give some initial feedback. Overall it looks great!

Also, if you haven't already, I would recommend testing this out with autoref as well! #2788

nimazareian · 2024-02-24T03:44:57Z

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.cpp

+    // Create move tactics
+    Vector positioning_vector = (world.ball().position() - placement_point);
+
+    // If the ball is nearly placed, then adjust move tactics to position between
+    // friendly goal and the ball
+    if (positioning_vector.length() < 0.5)
+    {
+        positioning_vector = world.field().friendlyGoalCenter() - world.ball().position();
+    }
+    positioning_vector = positioning_vector.normalize() * distance_to_keep;
+
+    Vector left_vector  = positioning_vector.rotate(Angle::fromDegrees(-30));
+    Vector right_vector = positioning_vector.rotate(Angle::fromDegrees(30));
+
+    Point center = world.ball().position() + positioning_vector;
+    Point left   = world.ball().position() + left_vector;
+    Point right  = world.ball().position() + right_vector;
+
+    move_tactics[0]->updateControlParams(
+        center, positioning_vector.orientation() + Angle::half(), 0);
+    move_tactics[1]->updateControlParams(left, left_vector.orientation() + Angle::half(),
+                                         0);
+    move_tactics[2]->updateControlParams(right,
+                                         right_vector.orientation() + Angle::half(), 0);
+    tactics_to_run[0].emplace_back(move_tactics[0]);
+    tactics_to_run[0].emplace_back(move_tactics[1]);
+    tactics_to_run[0].emplace_back(move_tactics[2]);


This would be interesting to see. We just gotta becareful of not really following the enemy robot placing the ball, as it could make it very easy to accidently get a foul if we get too close.

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.cpp

…into raymond/enemy_ball_placement

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_test.py

itsarune · 2024-04-01T08:23:39Z

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.cpp

+void EnemyBallPlacementPlayFSM::avoid(const Update& event)
+{
+    PriorityTacticVector tactics_to_run = {{}};
+
+    WorldPtr world_ptr = event.common.world_ptr;
+
+    // Create crease defenders
+    crease_defenders[0]->updateControlParams(placement_point,
+                                             TbotsProto::CreaseDefenderAlignment::LEFT);
+    crease_defenders[1]->updateControlParams(placement_point,
+                                             TbotsProto::CreaseDefenderAlignment::RIGHT);
+
+    tactics_to_run[0].emplace_back(crease_defenders[0]);
+    tactics_to_run[0].emplace_back(crease_defenders[1]);
+
+    // Create move tactics
+    Vector positioning_vector = (world_ptr->ball().position() - placement_point);
+
+    // If the ball is nearly placed, then adjust move tactics to position between
+    // friendly goal and the ball
+    if (positioning_vector.length() < 0.5)
+    {
+        positioning_vector = world_ptr->field().friendlyGoalCenter() - world_ptr->ball().position();
+    }
+    positioning_vector = positioning_vector.normalize() * distance_to_keep;
+
+    Vector left_vector  = positioning_vector.rotate(Angle::fromDegrees(-30));
+    Vector right_vector = positioning_vector.rotate(Angle::fromDegrees(30));
+
+    Point center = world_ptr->ball().position() + positioning_vector;
+    Point left   = world_ptr->ball().position() + left_vector;
+    Point right  = world_ptr->ball().position() + right_vector;
+
+    move_tactics[0]->updateControlParams(
+        center, positioning_vector.orientation() + Angle::half(), 0);
+    move_tactics[1]->updateControlParams(left, left_vector.orientation() + Angle::half(),
+                                         0);
+    move_tactics[2]->updateControlParams(right,
+                                         right_vector.orientation() + Angle::half(), 0);
+    tactics_to_run[0].emplace_back(move_tactics[0]);
+    tactics_to_run[0].emplace_back(move_tactics[1]);
+    tactics_to_run[0].emplace_back(move_tactics[2]);
+
+    // Set tactics to run
+    event.common.set_tactics(tactics_to_run);
+}


how about an alternate implementation?

we assign 6 move tactics

for each tactic, the robot stays in place unless it is inside the stadium obstacle

if it intersects the obstacle, we move to the closest point toward the centre of the field

That seems like an effective way of handling the edge case. However, I pulled from upstream and after some testing, I think Nima's #3152 PR fixes the issue of robots driving into the stadium obstacle, so the edge case of the goalie and crease defenders being in the stadium isn't really an issue anymore. I will work on adding tests for now, does that sound good?

Incorporated into the new implementation.

itsarune

nice work, looking good... left some comments about an alternate implementation that addresses your edge case that you found

…into raymond/enemy_ball_placement

…ides motion constraints

nimazareian

Just tried the new implementation with auto-ref and it looks great! I'm excited to get this in and use autoref more often now. 🟨

I just had a few nits.

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.h

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.cpp

src/software/simulated_tests/robot_enters_placement_region.py

…into raymond/enemy_ball_placement

raymond212

Addressed Nima's comments, new changes are ready for review.

nimazareian

LGTM! ⚽

Just had two additional non-blocking comments

nimazareian · 2024-05-25T07:11:49Z

src/proto/parameters.proto

@@ -37,6 +37,7 @@ message AiConfig
    required EnemyCapabilityConfig enemy_capability_config                  = 7;
    required ShootOrPassPlayConfig shoot_or_pass_play_config                = 8;
    required RobotCapabilitiesConfig robot_capabilities_config              = 9;
+    required EnemyBallPlacementPlayConfig enemy_ball_placement_play_config  = 10;


it's interesting that we had skipped field 10

nimazareian · 2024-05-25T07:19:35Z

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.cpp

+            // If either destination point is outside the field, then pick the other
+            // point. It is impossible for both points to be outside the field


Isn't it technically possible for both p1 and p2 (shown in red) to be outside of the field if the ball is in the corner and the placement position is somewhere towards the center? (the straight line shows the "backbone")

Though in either case we can probably ignore this edge case

…Software into raymond/enemy_ball_placement

williamckha

LGTM! Fouls begone

williamckha · 2024-05-27T06:15:17Z

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_fsm.h

+/**
+ * This FSM implements the enemy ball placement play.
+ * - If the ball placement point does not exist yet, then we wait
+ * - If the ball placement point exists, but the ball is far from the placement point,
+ * then we focus on avoiding ball placement interference. If our robot R is in the ball
+ * placement stadium, then it will move perpendicular to the segment connecting * and +,
+ * to either p1 or p2, whichever one is within the field boundaries and closest to R
+ *
+ *     /‾‾‾‾‾\
+ *     |  +  |        placement point
+ *     |     |
+ *     |     |
+ *  p1 |  R  | p2
+ *     |     |
+ *     |  *  |        ball
+ *     \_____/
+ *
+ * - Once the ball is near the placement point, we set up 1 goalie, 2 crease defenders and
+ * 3 robots to stay near the ball without interfering
+ *
+ *            +
+ *      R   x  enemy robot with ball
+ *        R
+ *          R
+ *
+ *      D   D
+ *
+ *        G
+ *     +-----+
+ */


Helpful diagrams!

raymond212 added 10 commits February 3, 2024 11:17

Increase shadow distance, improve transition into stop play

0a32ec1

Use constant for interference distance, fix shadowing side

94cea19

Merge branch 'master' of https://github.com/UBC-Thunderbots/Software …

cc01ee3

…into raymond/enemy_free_kick

Fix sensor fusion not setting ball placement position for enemy ball …

779bc10

…placement

Remove redundant helper function

496fbba

Use constant for stadium obstacle radius

03c7983

Modify enemy ball placement shadowing behaviour

d9bfea0

Fix formatting

b8b8a58

Convert enemy ball placement to use fsms

768c7c6

Small edits

b3ac962

raymond212 added the Gameplay label Feb 24, 2024

Fix formatting, add docs

39f3370

nimazareian reviewed Feb 24, 2024

View reviewed changes

raymond212 added 3 commits February 24, 2024 11:46

Use robot radius for distance to keep

abb490e

Merge branch 'master' of https://github.com/UBC-Thunderbots/Software …

efc392c

…into raymond/enemy_ball_placement

Update world to world_ptr

77a85e7

itsarune reviewed Apr 1, 2024

View reviewed changes

src/software/ai/hl/stp/play/enemy_ball_placement/enemy_ball_placement_play_test.py Outdated Show resolved Hide resolved

itsarune reviewed Apr 1, 2024

View reviewed changes

raymond212 added 10 commits April 4, 2024 14:52

Merge branch 'master' of https://github.com/UBC-Thunderbots/Software …

1831dfb

…into raymond/enemy_ball_placement

Add fsm diagram

1ccf439

Fix comments in fsm header

0ec21a9

Support stadium validation

8c5a572

Modify crease defenders

5479eb4

Wip simulated test

0708c71

Fix no placement point warning in sensor fusion

7df6a45

Remove commented out code

ee57406

Merge branch 'master' of https://github.com/UBC-Thunderbots/Software …

0793b93

…into raymond/enemy_ball_placement

Fix sensor fusion test

703da57

raymond212 requested a review from williamckha May 18, 2024 21:36

raymond212 and others added 6 commits May 18, 2024 15:20

Fix/delete ball placement motion contraint tests

ac02676

Revert motion constraint changes, use MoveTactic duplicate that overr…

e8bcac8

…ides motion constraints

[pre-commit.ci lite] apply automatic fixes

6a60364

Keep robots aligned to ball during avoid state

5ede280

Fix formatting

895f3ec

Use field lines instead of field boundary

9c01020

raymond212 requested a review from germchow May 21, 2024 14:26

nimazareian reviewed May 23, 2024

View reviewed changes

raymond212 added 4 commits May 23, 2024 21:55

Fix nits

aa2cc6f

Add avoid interference calculation docs, fix formatting

f545887

Use dynamic param for distance to keep

cd409d1

Merge branch 'master' of https://github.com/UBC-Thunderbots/Software …

219f3fe

…into raymond/enemy_ball_placement

raymond212 commented May 25, 2024

View reviewed changes

raymond212 requested a review from nimazareian May 25, 2024 04:59

[pre-commit.ci lite] apply automatic fixes

bdd4fa2

nimazareian previously approved these changes May 25, 2024

View reviewed changes

raymond212 added 2 commits May 25, 2024 11:52

Modify comment

15e8faa

Merge branch 'raymond/enemy_ball_placement' of github.com:raymond212/…

376ac66

…Software into raymond/enemy_ball_placement

raymond212 dismissed nimazareian’s stale review via 376ac66 May 25, 2024 18:53

nimazareian previously approved these changes May 25, 2024

View reviewed changes

Add safety margin for stop play robot speed

6d8d14f

raymond212 dismissed nimazareian’s stale review via 6d8d14f May 25, 2024 20:09

Fix formatting

396b53f

raymond212 requested a review from nimazareian May 25, 2024 20:52

williamckha approved these changes May 27, 2024

View reviewed changes

nimazareian approved these changes May 31, 2024

View reviewed changes

nimazareian mentioned this pull request Jun 1, 2024

Improve FreeKickPlay behaviour and convert it to use FSMs #2953

Merged

5 tasks

raymond212 merged commit 2555aaa into UBC-Thunderbots:master Jun 1, 2024
6 checks passed

raymond212 mentioned this pull request Jun 1, 2024

EnemyFreeKickPlay FSM Conversion #3179

Merged

4 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Improve EnemyBallPlacementPlay and convert it to use FSMs #3146

Improve EnemyBallPlacementPlay and convert it to use FSMs #3146

raymond212 commented Feb 24, 2024 •

edited

Loading

nimazareian commented Feb 24, 2024

nimazareian left a comment

nimazareian Feb 24, 2024

itsarune Apr 1, 2024

raymond212 Apr 4, 2024

raymond212 May 18, 2024

itsarune left a comment

nimazareian left a comment

raymond212 left a comment

nimazareian left a comment

nimazareian May 25, 2024

nimazareian May 25, 2024

williamckha left a comment

williamckha May 27, 2024

		// If either destination point is outside the field, then pick the other
		// point. It is impossible for both points to be outside the field

Improve EnemyBallPlacementPlay and convert it to use FSMs #3146

Improve EnemyBallPlacementPlay and convert it to use FSMs #3146

Conversation

raymond212 commented Feb 24, 2024 • edited Loading

Description

Testing Done

Resolved Issues

Length Justification and Key Files to Review

Review Checklist

nimazareian commented Feb 24, 2024

nimazareian left a comment

Choose a reason for hiding this comment

nimazareian Feb 24, 2024

Choose a reason for hiding this comment

itsarune Apr 1, 2024

Choose a reason for hiding this comment

raymond212 Apr 4, 2024

Choose a reason for hiding this comment

raymond212 May 18, 2024

Choose a reason for hiding this comment

itsarune left a comment

Choose a reason for hiding this comment

nimazareian left a comment

Choose a reason for hiding this comment

raymond212 left a comment

Choose a reason for hiding this comment

nimazareian left a comment

Choose a reason for hiding this comment

nimazareian May 25, 2024

Choose a reason for hiding this comment

nimazareian May 25, 2024

Choose a reason for hiding this comment

williamckha left a comment

Choose a reason for hiding this comment

williamckha May 27, 2024

Choose a reason for hiding this comment

raymond212 commented Feb 24, 2024 •

edited

Loading