Scoring is a subsystem inside the plugin, which is in charge of performing filter and order of perceived targets, eventually choosing one of those targets as the prioritary one inside the behavior tree.

Clicking behavior component that should be in your agent class, there are exposed settings for managing scoring, we will explain how to handle the different parameters below:

Scoring Method:

Currently (ver.1.0.0) we support only Manual, Automatic and EQS scoring methods, in the next updates the rest will be explained here:

Manual: Manual scoring will let you score the perceived actors of that affiliation, in order to pick your scored target you’ll need to inform the system, to do so, you must use Try update perceived target by affiliation:

EQS: We integrate Unreal EQS system with our plugin so you can define your custom EQS Tests, and thus choose a target defined in those tests. Once you choose the EQS method as your scoring method for a particular affiliation you can set the EnvQuery as shown in the image below:

Automatic: We defined a common Targetting method for perceived targets, which is parametrized in several segments, explained below:

Each test, while may differ from each other due to their innate differences, will have common settings, first we will explain Test Main Properties:

• Filter Mode: In case this targetting settings handles filtering, this mode will let targets pass if any filtering test allows it to pass, otherwise if AllPass is set as filter mode, then all filters must allow the target to pass to successfully allow it as a possible target.

• Test Mode: Each test in automatic method, will have a test mode. You can select between four options explained below:

  • Disable: This test won’t be executed at all

  • Filter only: This test will only filter the targets.

  • Score only: This test will only score the targets

  • Score and Filter: This test will filter and score targets

• Filtering Bounds: Each test will have tied bounds that can be used to filter the target based on the tested attribute, this is where you can choose how this filtering will behave.

• Scoring Settings: Each test will let you define scoring settings that will be used if the test itself is allowed to score the targets. Below are the parameters that you can customize to affect scoring:

  • Normalization Mode: How the score each target get will be normalized

    • Absolute: Lowest score a target can get will be 0, using this as a pivot, the scores will be ranged between 0 and the max value of the attribute being tested from all the tested actors.

    • RelativeToScores: Lowest and Max scores will be defined by the lowest and top scored targets respectively.

    • UserDefined: Normalization Ranges are User Defined.

  • UserDefined Bounds: Bound defined by user to be used for normalizing the targets’ score.

  • ScoreShaping|FunctionType: Function type to range the uniform score to get the shaped score of the targets.

  • ScoreShaping|Range: New Range of the uniform ranged score.

  • ScoreShaping|Factor: Factor used to scale the final score, if less than 0 it also inverts Score shaping range.

Perception based Tests: Most tests are perception based and these will have extra parameters based on the sense they are getting the data from.

• UsePerceptionData: Perception-based tests will have this parameter available, letting you to choose the perception sense to get data from to make the test.

• UseActorDataIfPerceptionIsNotFound: Whether to use tested actor’s location if perception based data is not found.

Independent Filters: Automatic Targetting lets you customize a range of filters as well.

• Use Line Sight Check: Lets you filter the tested actors based on whether the agent has line sight to the target or not.

• Perceived by Sense Check: Lets you filter the tested actors based on whether the target is currently perceived by a certain set of senses or not.

• Gameplay Tag Query Check: Lets you define a Gameplay tag query to filter the tested targets with. The tested targets must implement GameplayTag Interface.

Custom Attribute Tests: You can also implement tests based on attributes that are external to our systems, to do this first you’ll need to implement BPI_DynamicAICustomAttributes interface in your target’s class. By doing so, you can override CheckCustomAttribute Function and return a map of labeled attributes values.

After setting your custom attributes you can define the test as following:

And finally you can customize it as any other test.

Targetting Presets: Using automatic targetting method you can assign targetting preset, that will save you configuration time in case your desired targetting is found within those presets. Each preset comes with a comment that will let you know easily how the targetting will behave. To apply the presets you’ll need to press “Update Targetting Settings“ button or call it from BP.

In case you want to save your custom made targetting settings as a preset, you can copy and paste it in a new row of “DT_DynamicAI_TargettingPresets”

Score Run Mode:

The run mode will represent how you want the AI to select among the scored targets, the way it works is the same as EQS does with the run mode:

Single Best Item: Will choose as target the perceived target of that affiliation, with the highest score.

Single Random Item from Best 5%: Will choose randomly between the top 5% of the scored targets.

Single Random Item from Best 25%: Will choose randomly between the top 25% of the scored targets.

All Matching: Will choose randomly from all scored targets.

Scoring Delta:

This parameter will allow you to manage how often the scoring of the targets will take. Scoring the targets runs independently from the perception synchronization.

Target Relevancy Threshold:

This parameter lets you define whether the final selected target should be considered for active behavior treatment, this means that after the target scoring finish and one target is selected, the final score that it holds can be tested against this threshold value, if its greater, a certain behavior can be chosen to run, otherwise a different behavior can be executed.

A good example is found in the combat template behavior, If the score of the selected target is greater than the considered Active Target Threshold for that affiliation, the agent will try to engaged in combat with the target, if its lesser it will trigger investigation sub behavior.

Scoring Debug:

By enabling debug, you can see at run time how your agent scores his targets.