[Gameplay] Avoid Aimless Kick/Chip by Defenders by Andrewyx · Pull Request #3230 · UBC-Thunderbots/Software

Andrewyx · 2024-06-22T08:59:56Z

Please fill out the following before requesting review on this PR

Description

Changes Crease Defender Autochip behaviour such that robots only chip in dire circumstances.
Further adds new Chip Validation

creasechip.mp4

Crease defenders now only autochip incoming balls when:

Crease Defender is not facing friendly net
Crease Defender has enemy in front to justify chipping over
The ball && crease defender is near the friendly net

Further adds the following:

Also gives crease defenders the ability to chase after ball with DribbleFSM when the ball is nearby and unguarded based on threshold values.
Includes new crease_defender_tactic_test.py pytest with simulated tests. Removed/replaces certain outdated C++ tests.

Testing Done

Adds new chip validation test which passes when ball is threshold meters off the ground. This allows us to test if chip behaviour occurs.

Resolved Issues

Resolves #3220

Length Justification and Key Files to Review

This PR includes a new Simulated Test along with a new Validation Type.

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

Suggests new ticket reworking the visual component of the ball_if_off_ground validation test. This test validates changes in the z axis and the current 2D shape system is not enough to visualize this. It is purely a visual change and is not urgent, but will take some time to think of a new way to visualize.

src/software/ai/hl/stp/stp_tactic_assignment_test.cpp

itsarune · 2024-06-26T04:00:50Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+    std::vector<Robot> enemy_robots = event.common.world_ptr->enemyTeam().getAllRobots();
+    for (int i = 0; i < static_cast<int>(event.common.world_ptr->enemyTeam().numRobots()); i++) {
+        if (contains(zone, enemy_robots[i].position()))
+        {
+            return true;
+        }
+    }
+    return false;


nit: try refactoring to use std::any_of: https://en.cppreference.com/w/cpp/algorithm/all_any_none_of

itsarune · 2024-06-26T04:02:00Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+    // DEBUG: Visualizes threat stadium
+    LOG(VISUALIZE) << *createDebugShapes({
+         *createDebugShape(threat_zone, std::to_string(event.common.robot.id()), "threatzone")
+    });


remove to delete when you're done

itsarune · 2024-06-26T04:03:16Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+
+    if (goal_intersections.empty()
+        && CreaseDefenderFSM::isAnyEnemyInZone(event, threat_zone)
+        && robot_to_net_m <= event.common.world_ptr->field().totalYLength() / 2)


why did you use YLength() here?

also this should be a constant with the word THRESHOLD in it

This ensures that chips only occur when the ball is somewhat near our net. This cannot be a constant since this value must scale with field size and this is only determined in runtime

itsarune · 2024-06-26T04:04:14Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+    if (goal_intersections.empty()
+        && CreaseDefenderFSM::isAnyEnemyInZone(event, threat_zone)
+        && robot_to_net_m <= event.common.world_ptr->field().totalYLength() / 2)
+    {
+        // Autochip only if the robot is not facing the net, there is an enemy in front, and robot is close to net
+        auto_chip_or_kick = AutoChipOrKick{
+            AutoChipOrKickMode::AUTOCHIP,
+            chip_distance
+        };
+    }
+    else
+    {
+        auto_chip_or_kick = AutoChipOrKick{AutoChipOrKickMode::OFF, 0};
+    }


Suggested change

if (goal_intersections.empty()

&& CreaseDefenderFSM::isAnyEnemyInZone(event, threat_zone)

&& robot_to_net_m <= event.common.world_ptr->field().totalYLength() / 2)

{

// Autochip only if the robot is not facing the net, there is an enemy in front, and robot is close to net

auto_chip_or_kick = AutoChipOrKick{

AutoChipOrKickMode::AUTOCHIP,

chip_distance

};

}

else

{

auto_chip_or_kick = AutoChipOrKick{AutoChipOrKickMode::OFF, 0};

}

auto_chip_or_kick = AutoChipOrKick{AutoChipOrKickMode::OFF, 0};

if (goal_intersections.empty()

&& CreaseDefenderFSM::isAnyEnemyInZone(event, threat_zone)

&& robot_to_net_m <= event.common.world_ptr->field().totalYLength() / 2)

{

// Autochip only if the robot is not facing the net, there is an enemy in front, and robot is close to net

auto_chip_or_kick = AutoChipOrKick{

AutoChipOrKickMode::AUTOCHIP,

chip_distance

};

}

itsarune · 2024-06-26T04:39:15Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+    constexpr double GET_POSSESSION_THRESHOLD_M = 1;
+    constexpr  double THREAT_THRESHOLD_M = GET_POSSESSION_THRESHOLD_M * 2;


this should be in the header so that we can find these constants easily

itsarune · 2024-06-26T04:39:39Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+    LOG(VISUALIZE) << *createDebugShapes({
+             *createDebugShape(
+                     Circle(robot_position, GET_POSSESSION_THRESHOLD_M),
+                     std::to_string(event.common.robot.id()) + "1",
+                     "ballzone"
+                     )
+     });


remember to remove this when you're done

itsarune · 2024-06-26T04:40:50Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+                                             boost::sml::back::process<DribbleFSM::Update> processEvent)
+{
+    DribbleFSM::ControlParams control_params{
+        .dribble_destination = Point(),


Suggested change

.dribble_destination = Point(),

.dribble_destination = event.common.world_ptr->ball().position(),

itsarune · 2024-06-26T04:41:09Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

+{
+    DribbleFSM::ControlParams control_params{
+        .dribble_destination = Point(),
+        .final_dribble_orientation = Angle(),


should be oriented towards the enemy net

itsarune · 2024-06-26T04:42:19Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.h

-            *MoveFSM_S + Update_E / blockThreat_A, MoveFSM_S = X,
+            *MoveFSM_S + Update_E[ballNearbyWithoutThreat_G] / prepareGetPossession_A = DribbleFSM_S,
+            MoveFSM_S + Update_E / blockThreat_A, MoveFSM_S = X,
+            DribbleFSM_S + Update_E[!ballNearbyWithoutThreat_G] / blockThreat_A = X,


Suggested change

DribbleFSM_S + Update_E[!ballNearbyWithoutThreat_G] / blockThreat_A = X,

DribbleFSM_S + Update_E[!ballNearbyWithoutThreat_G] / blockThreat_A = MoveFSM,

itsarune · 2024-06-26T04:42:37Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.h

+     *
+     * @param event CreaseDefenderFSM::Update event
+     * @param zone a stadium shape that defines the zone
+     * @param zone a stadium shape that defines the zone


Suggested change

* @param zone a stadium shape that defines the zone

itsarune · 2024-06-26T04:43:30Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic.cpp

+    fsm_map.at(tactic_update.robot.id())
+        ->process_event(DribbleFSM::Update(dribble_control_params, tactic_update));


do we need this? I'm not 100% sure if this is necessary

No idea why, but weirdly yes

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic.h

src/software/ai/hl/stp/tactic/tactic_factory.cpp

itsarune

left some interim feedback. Looks like one of the CHECKs fail when we play AI vs AI. I think the update that I suggested to crease_defender_fsm.h may fix this issue

…nto Andrewyx/AimlessChip

itsarune · 2024-07-08T01:16:47Z

src/proto/parameters.proto

+     */
+    required double max_get_ball_ratio_threshold = 1
+        [default = 0.3, (bounds).min_double_value = 0.0, (bounds).max_double_value = 1.0];
+    // Max distance that the crease will try and get possession of a ball


Suggested change

// Max distance that the crease will try and get possession of a ball

// Max distance that the crease defenderwill try and get possession of a ball

itsarune · 2024-07-08T01:27:53Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.h

        const Field& field, const Ray& ray, double robot_obstacle_inflation_factor);

-   private:
+    /**


still should be private:

Oh i see why you made it public. The class members should still be private though

Oh yeah, this part is a bit deceptive, I just moved the private up. These functions are actually private, which I suppose means theres no point making them static.

https://github.com/UBC-Thunderbots/Software/pull/3230/files/fb4fd16c47a2bc06bd1f4d7a4782d865e7dc0d06#diff-1fe03c39bfde2b5406fa4e07f3683bf5791135da7b6ab57dd021af3100c97d7dR112-R114

itsarune · 2024-07-08T01:36:37Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+friendlyDefenseAreaFrontCenter = tbots_cpp.Point(
+    tbots_cpp.Field.createSSLDivisionBField()
+    .friendlyDefenseArea()
+    .posXPosYCorner()
+    .x(),
+    tbots_cpp.Field.createSSLDivisionBField().friendlyDefenseArea().centre().y(),
+)


this is just tbots_cpp.Field.createSSLDivisionBField().friendlyDefenseArea().centre() btw

The point I want is the center of the frontal segment of the defence area, not necessarily the exact center of the defense box.

itsarune · 2024-07-08T01:37:24Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+        # These aren't necessary for this test, but this is just an example
+        # of how to send commands to the simulator.
+        #
+        # NOTE: The gamecontroller responses are automatically handled by
+        # the gamecontroller context manager class
+        simulated_test_runner.gamecontroller.send_gc_command(
+            gc_command=Command.Type.STOP, team=Team.UNKNOWN
+        )
+        simulated_test_runner.gamecontroller.send_gc_command(
+            gc_command=Command.Type.FORCE_START, team=Team.BLUE
+        )


you can delete this, I don't believe it's necessary

itsarune · 2024-07-08T01:37:42Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+        # These aren't necessary for this test, but this is just an example
+        # of how to send commands to the simulator.
+        #
+        # NOTE: The gamecontroller responses are automatically handled by
+        # the gamecontroller context manager class
+        simulated_test_runner.gamecontroller.send_gc_command(
+            gc_command=Command.Type.STOP, team=Team.UNKNOWN
+        )
+        simulated_test_runner.gamecontroller.send_gc_command(
+            gc_command=Command.Type.FORCE_START, team=Team.BLUE
+        )


you can delete these lines, it's probably not necessary

itsarune · 2024-07-08T01:39:35Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+        # These aren't necessary for this test, but this is just an example
+        # of how to send commands to the simulator.
+        #
+        # NOTE: The gamecontroller responses are automatically handled by
+        # the gamecontroller context manager class
+        simulated_test_runner.gamecontroller.send_gc_command(
+            gc_command=Command.Type.STOP, team=Team.UNKNOWN
+        )
+        simulated_test_runner.gamecontroller.send_gc_command(
+            gc_command=Command.Type.FORCE_START, team=Team.BLUE
+        )


you can delete this

itsarune · 2024-07-08T01:41:00Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+        params.assigned_tactics[0].crease_defender.CopyFrom(
+            CreaseDefenderTactic(
+                enemy_threat_origin=tbots_cpp.createPointProto(ball_initial_pos),
+                crease_defender_alignment=2,


you can probably do CreaseDefenderAlignment.CENTRE here

itsarune · 2024-07-08T01:41:27Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+        params.assigned_tactics[0].crease_defender.CopyFrom(
+            CreaseDefenderTactic(
+                enemy_threat_origin=tbots_cpp.createPointProto(ball_initial_pos),
+                crease_defender_alignment=2,


you can probably do CreaseDefenderAlignment.CENTRE here

itsarune · 2024-07-08T01:41:37Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+        params.assigned_tactics[0].crease_defender.CopyFrom(
+            CreaseDefenderTactic(
+                enemy_threat_origin=tbots_cpp.createPointProto(ball_initial_pos),
+                crease_defender_alignment=2,


you can probably do CreaseDefenderAlignment.CENTRE here

itsarune · 2024-07-08T01:41:55Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

great tests!

itsarune · 2024-07-08T01:42:25Z

src/software/simulated_tests/ball_is_off_ground.py

+
+
+class BallIsOffGround(Validation):
+


nit:

Suggested change

…nto Andrewyx/AimlessChip

nimazareian

some nits

nimazareian · 2024-07-08T07:35:47Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

+    )
+
+
+friendlyDefenseAreaFrontCenter = tbots_cpp.Point(


nit: should be snake_case

nimazareian · 2024-07-08T07:41:15Z

src/proto/parameters.proto

+    Threshold = 1.0 results in crease defenders chasing the ball even ball is equidistant
+    to nearest enemy. Threshold < 1.0 results in crease defenders chasing the ball only
+    when it is at least (threshold) percent closer to our bot than the nearest enemy.


nit:

Suggested change

Threshold = 1.0 results in crease defenders chasing the ball even ball is equidistant

to nearest enemy. Threshold < 1.0 results in crease defenders chasing the ball only

when it is at least (threshold) percent closer to our bot than the nearest enemy.

Threshold = 1.0 results in crease defenders chasing the ball even if the ball is equidistant

to nearest enemy. Threshold < 1.0 results in crease defenders chasing the ball only

when it is at least (threshold) percent closer to our bot than the nearest enemy.

nimazareian · 2024-07-08T07:45:58Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

        bool ball_is_near_friendly =
-            ball_distance < nearest_enemy_distance *
-                                crease_defender_config.max_get_ball_ratio_threshold();
+            ball_distance_to_friendly <
+            ball_distance_to_enemy *
+                (1.0 - crease_defender_config.max_get_ball_ratio_threshold());


This change in the logic seems backwards to me. Shouldn't we be multiplying by crease_defender_config.max_get_ball_ratio_threshold() directly to get the ratio?

You are right that mathematically speaking, it would be a direct multiplication. However, I think the meaning of max_get_ball_ratio_threshold() makes a lot more sense when we flip the percentage. For example, if max_get_ball_ratio_threshold() = 0.1, this means that the ball must be at least 10% closer to our defender than the enemy for stealing to occur. Within the code though, this could be either represented by a multiplication of 0.9 to the enemy dist, or a multiplication of 1.1 to friendly dist.

Below is some rambling since I changed the meaning of the math here today. It is only if you are curious why I made a change here.

I thought of an edge case and switched the logic to use this instead. The old code directly multiplied it since it was comparing the distance of the defender->ball vs nearest_enemy_to_defender->ball. This works in most circumstances, but in the case where the world looks like this:
Ball Crease Enemy
x O O
The crease would decide not to pursue the ball when it really should. Hence, I changed the logic to instead measure the distance from the ball to the nearest enemy and compare that with the current crease's distance.

This means that I redefined the meaning of max_get_ball_ratio_threshold()

/* Max ratio between distances (crease defender and ball) / (nearest enemy and ball) for pass defender to chase ball. Threshold = 1.0 results in crease defenders chasing the ball even ball is equidistant to nearest enemy. Threshold < 1.0 results in crease defenders chasing the ball only when it is at least (threshold) percent closer to our bot than the nearest enemy. */

I followed up in a new comment. Though Im still not sure if I fully understand the logic. I think I understand the edge case you're talking about, but I think in that case it's also okay to act as a regular crease defender trying to block the ball, as opposed to trying to take a risk by attempting to steal the ball.

Also, regarding the comment:

Threshold = 1.0 results in crease defenders chasing the ball even ball is equidistant to nearest enemy. Threshold < 1.0 results in crease defenders chasing the ball only when it is at least (threshold) percent closer to our bot than the nearest enemy.

If the value is set to 1, then right side of the equality becomes 0, meaning that the ball distance to friendly has to be less than 0 (impossible) for the friendly to try to steal the ball. That's why I think just multiplying makes more sense. Threshold of 1 should indeed mean that the crease defender is the most aggressive. And 0 be that the crease defender doesn't take any risks.

nimazareian · 2024-07-08T07:57:50Z

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

@@ -30,10 +34,21 @@ std::optional<Point> CreaseDefenderFSM::findBlockThreatPoint(
    return findDefenseAreaIntersection(field, ray, robot_obstacle_inflation_factor);


You should make findDefenseAreaIntersection static so findBlockThreatPoint can call it

Ohhhh right, that's why I originally wrote it as static

src/proto/parameters.proto

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp

williamckha

💯

Andrewyx added 4 commits June 20, 2024 02:54

init commit

db7f5ce

changes

bab89e0

small update

97bc01d

changes

6275d9f

Andrewyx changed the title ~~Andrewyx/aimless chip~~ [Gameplay] Avoid Aimless Kick/Chip by Defenders Jun 22, 2024

Andrewyx added 3 commits June 22, 2024 13:33

ai config

7778f59

dribble fsm added

047b8cf

merged

311cedb

itsarune reviewed Jun 26, 2024

View reviewed changes

src/software/ai/hl/stp/stp_tactic_assignment_test.cpp Show resolved Hide resolved

itsarune reviewed Jun 26, 2024

View reviewed changes

tests changed

de9eadf

itsarune reviewed Jun 26, 2024

View reviewed changes

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic.h Show resolved Hide resolved

itsarune reviewed Jun 26, 2024

View reviewed changes

src/software/ai/hl/stp/tactic/tactic_factory.cpp Show resolved Hide resolved

itsarune reviewed Jun 26, 2024

View reviewed changes

Andrewyx added 6 commits June 26, 2024 00:24

test fixed

87306b4

init test

652dbf0

working test

fec1a06

working tests

c28d6e4

new validation added

b094e99

basic tests added

65d7e44

Andrewyx and others added 3 commits July 7, 2024 14:53

Merge branch 'Andrewyx/AimlessChip' of github.com:Andrewyx/Software i…

1c09358

…nto Andrewyx/AimlessChip

nits addressed

ecd2f6b

[pre-commit.ci lite] apply automatic fixes

fb4fd16

itsarune reviewed Jul 8, 2024

View reviewed changes

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_tactic_test.py

Copy link

Contributor

itsarune Jul 8, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

great tests!

itsarune reviewed Jul 8, 2024

View reviewed changes

Andrewyx added 2 commits July 7, 2024 23:52

nits fixed

1c773e1

Merge branch 'Andrewyx/AimlessChip' of github.com:Andrewyx/Software i…

2f48a3f

…nto Andrewyx/AimlessChip

nimazareian reviewed Jul 8, 2024

View reviewed changes

Andrewyx and others added 2 commits July 8, 2024 01:31

fixes

81d1108

[pre-commit.ci lite] apply automatic fixes

e73a8b3

Andrewyx requested review from itsarune, nimazareian and williamckha July 8, 2024 17:27

Merge branch 'master' into Andrewyx/AimlessChip

59ef1b7

nimazareian added the Robocup-2024 label Jul 9, 2024

nimazareian reviewed Jul 9, 2024

View reviewed changes

src/proto/parameters.proto Show resolved Hide resolved

src/software/ai/hl/stp/tactic/crease_defender/crease_defender_fsm.cpp Outdated Show resolved Hide resolved

fixes

caab2c6

nimazareian approved these changes Jul 10, 2024

View reviewed changes

williamckha approved these changes Jul 10, 2024

View reviewed changes

Andrewyx merged commit 00372df into UBC-Thunderbots:master Jul 10, 2024

		constexpr double GET_POSSESSION_THRESHOLD_M = 1;
		constexpr double THREAT_THRESHOLD_M = GET_POSSESSION_THRESHOLD_M * 2;

	.dribble_destination = Point(),
	.dribble_destination = event.common.world_ptr->ball().position(),

	DribbleFSM_S + Update_E[!ballNearbyWithoutThreat_G] / blockThreat_A = X,
	DribbleFSM_S + Update_E[!ballNearbyWithoutThreat_G] / blockThreat_A = MoveFSM,

		fsm_map.at(tactic_update.robot.id())
		->process_event(DribbleFSM::Update(dribble_control_params, tactic_update));

	// Max distance that the crease will try and get possession of a ball
	// Max distance that the crease defenderwill try and get possession of a ball

		@@ -30,10 +34,21 @@ std::optional<Point> CreaseDefenderFSM::findBlockThreatPoint(
		return findDefenseAreaIntersection(field, ray, robot_obstacle_inflation_factor);

		)


		friendlyDefenseAreaFrontCenter = tbots_cpp.Point(

Conversation

Andrewyx commented Jun 22, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Please fill out the following before requesting review on this PR

Description

Testing Done

Resolved Issues

Length Justification and Key Files to Review

Review Checklist

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!

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!

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!

Uh oh!

Uh oh!

itsarune 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!

Andrewyx Jul 8, 2024 • 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!

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!

nimazareian 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

Andrewyx commented Jun 22, 2024 •

edited

Loading

Andrewyx Jul 8, 2024 •

edited

Loading