Public Member Functions
	NormalizedStallHeuristic (float minimumLinearError=0.01f, float directionSimilarityThreshold=0.95f, float correctionFractionThreshold=0.05f, float errorRatioThreshold=0.95f)
float	StalledErrorCorrectionDetection (NetworkTransform networkTransform, float physicsDt, AbstractPhysicsBody physicsBody, Vector3 previousLocalPosition, Vector3 previousExtrapolatedPosition, ref KinematicSnapshot localSnapshot, ref KinematicSnapshot extrapolatedSnapshot)
	Implements the NetworkTransform.CustomErrorDetectionDelegate so it can be provided as a error detection heuristic for use on forecasted NetworkTransforms.

Detailed Description

Error detection heuristic that detects persistent stuck states. differs from the standard implementation in that it is normalized and time-step independent and tuned via 3 fractional or precentage based thresholds

Evaluates three conditions:

Error direction similarity - are current and previous errors pointing the same way?
Correction progress - are we making progress closing the gap?
Error persistence - is the error magnitude staying the same or growing?

When all three indicate a problem (consistent error direction, no progress, error not shrinking), returns a positive score indicating error. Otherwise returns a negative score indicating healthy correction.

Constructor & Destructor Documentation

◆ NormalizedStallHeuristic()

NormalizedStallHeuristic	(	float	minimumLinearError = 0::01f,
		float	directionSimilarityThreshold = 0::95f,
		float	correctionFractionThreshold = 0::05f,
		float	errorRatioThreshold = 0::95f )

Constructs a NormalizedStallHeuristic

Once an instance has been made the NormalizedStallHeuristic.Delegate property can be assigned to the NetworkTransform.CustomErrorDetection delegate on a network transform to override the default stall heurstics algorithm used

Parameters

minimumLinearError	Error distances below this value will return -1f, the error is too small to care
directionSimilarityThreshold	What proportion of the error from the previous frame needs to match this frame to be considered stalled?
correctionFractionThreshold	What proportion of last frames error needs to have been corrected this frame to be considered not stalled?
errorRatioThreshold	How much does this frames error need to have shrunk by, compared with last frames error to not be considered stalled?

Member Function Documentation

◆ StalledErrorCorrectionDetection()

float StalledErrorCorrectionDetection	(	NetworkTransform	networkTransform,
		float	physicsDt,
		AbstractPhysicsBody	physicsBody,
		Vector3	previousLocalPosition,
		Vector3	previousExtrapolatedPosition,
		ref KinematicSnapshot	localSnapshot,
		ref KinematicSnapshot	extrapolatedSnapshot )

Implements the NetworkTransform.CustomErrorDetectionDelegate so it can be provided as a error detection heuristic for use on forecasted NetworkTransforms.