CN100479360C - Restraining method of control points of using marginal effect - Google Patents

Abstract

Based on moving speed, acceleration of object, and distance of dragged control point, the invention adjusts predicted result dynamically. Based on distance between position of initial dragged control point of user and initial predicted position, the invention adjusts predicted weight. The more faraway the distance is, the more obvious of restriction is. Restriction is amplified in exponential. The invention compensates unfavored result brought by wobble of network, restricts prediction of movement of control point so as to raise precision of prediction for movement of control point, and reach better effect of prediction.

Description

Adopt the control point constrained procedure of marginal effect

Technical field

The present invention relates to the collaborative perception field of computer supported cooperative work, particularly relate to a kind of control point constrained procedure that adopts marginal effect.

Background technology

One of 20th century mankind's outstanding achievement computer technology has been brought human society into the information age.Be accompanied by deepening continuously of IT application process, the communication technology, computer and network technology merge mutually, have produced a new research field-computer supported cooperative work CSCW (Computer SupportedCooperative Work).

As collaborative design person during in co-operation, the monitoring Long-distance Control locus of points can help the user to understand preferably and predict other designers' intention.The change procedure that represents long-range pattern objects is for supporting that collaborative work is of great value means.Yet usually there is the instability of delay in network, also is called shake (jitter), and this may cause representing under the real time environment the inaccurate of remote participant activity.Shake can cause the discontinuous of Long-distance Control point motion track, causes the user to misread or even operating collision.In order to relax the negative effect brought of shake, adopt usually based on known historical position and predict that Long-distance Control puts the method for next position, with the track of skimulated motion, reach level and smooth natural effect.

There is certain limitation in track following (Tracing) technology at present, the main mutual effect that adopts the Dead-reckoning method to improve player and distributed objects in the Internet game on-line.The Dead-reckoning method has been utilized speed, the acceleration parameter of object, shows good effect in the moving process based on the strong inertial mass of mechanics.The Dead-reckoning Forecasting Methodology is applied to equally cuts down the shake effect that move at the control point, experimental results show that this prediction can improve in the group system immediacy and naturality mutual between the user.

Yet when the shake time was longer, the prediction accuracy of Dead-reckoning under this is used can obviously descend.Its reason is that generally the designer only can pay close attention to the editing area of fixing around it in a period of time.When the control point moved to other positions, new position was more approaching on distance with initial position probably.And the possibility that user's control point moves to outside the screen is also little, and the Dead-reckoning method is not considered this effect that marginalizes.

Summary of the invention

The object of the present invention is to provide a kind of control point constrained procedure that adopts marginal effect.

The technical scheme that the present invention solves its technical problem employing is as follows:

1) the Long-distance Control point of the nearest N of record within second is in the shift position of each millisecond;

2) when network jitter takes place, for the shift position of PREDICTIVE CONTROL point, at first calculation control point is used aveVelocity at the present speed of directions X and Y direction _XAnd aveVelocity _YRepresent; Approximate velocity is by from current nearest T _LastPosition coordinates (X constantly _Last, Y _Last) and the preceding T of M millisecond _FlagPosition coordinates (the X that constantly receives _Flag, Y _Flag) calculate;

$\left\{\begin{array}{c}\mathrm{aveVelocit}{y}_X=(X_{\mathrm{last}}-X_{\mathrm{flag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\\ \mathrm{aveVelocit}{y}_Y=(Y_{\mathrm{last}}-Y_{\mathrm{flag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\end{array}\right.$

3) calculation control point is used aveAcceleration in the current acceleration of directions X and Y direction _XAnd aveAcceleration _YRepresent; Approximate acceleration is by from current nearest T _Last(X constantly _Last, Y _Last) speed and the T of position _Flag(X constantly _Flag, Y _Flag) another speed of position calculates;

$\left\{\begin{array}{c}\mathrm{aveAcceleratio}{n}_X=({\mathrm{Velocity}}_{\mathrm{Xlast}}-{\mathrm{Velocity}}_{\mathrm{Xflag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\\ \mathrm{aveAcceleratio}{n}_Y=({\mathrm{Velocity}}_{\mathrm{Ylast}}-V{\mathrm{elocity}}_{\mathrm{Yflag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\end{array}\right.$

4) calculate the initial predicted position

The current location note at control point is (X _Current, Y _Current), and the next position note at control point is X _Next, Y _Next), when network jitter takes place, (X _Next, Y _Next) prediction calculate by following formula:

$\left\{\begin{array}{c}{X}_{\mathrm{next}}=X_{\mathrm{current}}+\mathrm{aveVelocit}{y}_X*T+\mathrm{aveAccelerati}{\mathrm{on}}_X*T^2\\ Y_{\mathrm{next}}=Y_{\mathrm{current}}+\mathrm{ave}{\mathrm{Velocity}}_Y*T+\mathrm{aveAcceleratio}{n}_Y*T^2\end{array}\right.$

Wherein:

T is current shake time-delay;

5) calculate the prediction weight

Initially drag the position at control point and the distance between the initial predicted position according to the user, calculate the prediction weight; Distance is far away more, retrains obvious more;

$\mathrm{\&alpha;}=\left\{\begin{array}{cc}1& ,\mathrm{Length}\&le;\mathrm{UnitLength}\\ {(\mathrm{Length}/\mathrm{UnitLength})}^b,0<b<1& ,\mathrm{Length}>\mathrm{UnitLength}\end{array}\right.$

Wherein:

Length is that the user initially drags the position at control point and the distance between the initial predicted position;

UnitLength can decide according to different systems and application;

B is greater than 0 constant less than 1;

The position that is initially dragged by the user is the center of circle, and UnitLength is that radius forms a scope, if the position of initial predicted in this scope, system keeps the position of original prediction not change, i.e. α=1; When the position of initial prediction exceeds this scope, system adopts (Length/UnitLength) ^bAs weight;

6) carry out the constraint that predict at the control point

Utilize the prediction weight to the initial predicted result constraint that further marginalizes, the initial predicted position is (X _Next, Y _Next), the control point marginalize the result note of constraint do (X ' _Next, Y ' _Next),

$\left\{\begin{array}{c}{X^{\&prime;}}_{\mathrm{next}}=X_{\mathrm{next}}+\mathrm{UnitLength}*\mathrm{\&alpha;}*\sin \mathrm{\&beta;}\\ {Y^{\&prime;}}_{\mathrm{next}}=Y_{\mathrm{next}}+\mathrm{UnitLength}*\mathrm{\&alpha;}*\cos \mathrm{\&beta;}\end{array}\right.$

Wherein:

Sin β represents the component of control point in the x direction;

Cos β represents the component of control point in the y direction;

7) with Long-distance Control point according to marginalize the constraint the result (X ' _Next, Y ' _Next) be plotted on user's the screen;

8) wipe the control point of previous prediction, draw the control point of up-to-date prediction, the rest may be inferred, and complete in the control point of all predictions in dither process.

The present invention compares with background technology, and the useful effect that has is:

Traditional Forecasting Methodology is not considered the effect that marginalizes, and the movement locus at dynamic prediction of the present invention control point initially drags the position at control point and the distance between the initial predicted position according to the user, adjusts the prediction weight.Far away more from initial position, the constraint that marginalizes is obvious more, and constraint presents index and strengthens.This prediction result more meets human operating habit.The present invention can be applied in the real-time perception problem of working in coordination with fields such as design preferably.

Embodiment

The specific implementation flow process of the control point constrained procedure of employing marginal effect is as follows.

1) the Long-distance Control point of the nearest N of record within second is in the shift position of each millisecond;

For example, can write down position data within nearest 2 seconds.

2) when network jitter takes place, for the shift position of PREDICTIVE CONTROL point, at first calculation control point is used aveVelocity at the present speed of directions X and Y direction _XAnd aveVelocity _YRepresent; Approximate velocity is by from current nearest T _LastPosition coordinates (X constantly _Last, Y _Last) and the preceding T of M millisecond _FlagPosition coordinates (the X that constantly receives _Flag, Y _Flag) calculate;

$\left\{\begin{array}{c}\mathrm{aveVelocit}{y}_X=(X_{\mathrm{last}}-X_{\mathrm{flag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\\ \mathrm{aveVelocit}{y}_Y=(Y_{\mathrm{last}}-Y_{\mathrm{flag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\end{array}\right.$

For example, the M value is got 100 milliseconds, promptly 0.1 second.T _LastPosition coordinates constantly is (310,320), T _FlagThe position coordinates that constantly receives is (300,300), coordinate all with pixel as unit.

$\left\{\begin{array}{c}\mathrm{aveVelocit}{y}_X=(X_{\mathrm{last}}-X_{\mathrm{flag}})/100=(310-300)/0.1=100\\ \mathrm{aveVelocit}{y}_Y=(Y_{\mathrm{last}}-Y_{\mathrm{flag}})/100=(320-300)/0.1=200\end{array}\right.$ (pixel/s)

3) calculation control point is used aveAcceleration in the current acceleration of directions X and Y direction _XAnd aveAcceleration _YRepresent; Approximate acceleration is by from current nearest T _Last(X constantly _Last, Y _Last) speed and the T of position _Flag(X constantly _Flag, Y _Flag) another speed of position calculates;

$\left\{\begin{array}{c}\mathrm{aveAcceleratio}{n}_X=({\mathrm{Velocity}}_{\mathrm{Xlast}}-\mathrm{Velocit}{y}_{\mathrm{Xflag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\\ \mathrm{ave}{\mathrm{Acceleration}}_Y=({\mathrm{Velocity}}_{\mathrm{Ylast}}-V{\mathrm{elocity}}_{\mathrm{Yflag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\end{array}\right.$

Suppose T _Last(X constantly _Last, Y _Last) speed of position is (100,100), acceleration calculation is as follows so:

$\left\{\begin{array}{c}\mathrm{aveAcceleratio}{n}_X=(100-100)/0.1=0\\ \mathrm{aveAcceleratio}{n}_Y=(200-100)/0.1=1000\end{array}\right.$ (pixel/s ²)

4) calculate the initial predicted position

The current location note at control point is (X _Current, Y _Current), and the next position note at control point is (X _Next, Y _Next), when network jitter takes place, (X _Next, Y _Next) prediction calculate by following formula:

$\left\{\begin{array}{c}{X}_{\mathrm{next}}=X_{\mathrm{current}}+\mathrm{aveVelocit}{y}_X*T+\mathrm{aveAccelerati}{\mathrm{on}}_X*T^2\\ Y_{\mathrm{next}}=Y_{\mathrm{current}}+\mathrm{ave}{\mathrm{Velocity}}_Y*T+\mathrm{aveAcceleratio}{n}_Y*T^2\end{array}\right.$

Wherein:

T is current shake time-delay;

As above example is supposed T=100 (ms)

$\left\{\begin{array}{c}{X}_{\mathrm{next}}=310+100*0.1=320\\ Y_{\mathrm{next}}=320+200*0.1+1000*{0.1}^2=350\end{array}\right.$

5) calculate the prediction weight

Initially drag the position at control point and the distance between the initial predicted position according to the user, calculate the prediction weight.Distance is far away more, retrains obvious more;

$\mathrm{\&alpha;}=\left\{\begin{array}{cc}1& ,\mathrm{Length}\&le;\mathrm{UnitLength}\\ {(\mathrm{Length}/\mathrm{UnitLength})}^b,0<b<1& ,\mathrm{Length}>\mathrm{UnitLength}\end{array}\right.$

Wherein:

Length is that the user initially drags the position at control point and the distance between the initial predicted position.

UnitLength can decide according to different systems and application.

B is greater than 0 constant less than 1;

The position that is initially dragged by the user is the center of circle, and UnitLength is that radius forms a scope.If the position of initial predicted is in this scope, system keeps the position of original prediction not change, i.e. α=1.When the position of initial prediction exceeds this scope, system adopts (Length/UnitLength) ^bAs weight.

For example, Length=200 pixel, UnitLength=100 pixel, b=1/2.Because Length＞UnitLength,

$\mathrm{\&alpha;}={(\mathrm{Length}/\mathrm{UnitLength})}^b={(200/100)}^{\frac{1}{2}}=2^{\frac{1}{2}}$

6) carry out the constraint that predict at the control point

Utilize the prediction weight to the initial predicted result constraint that further marginalizes.The initial predicted position is (X _Next, Y _Next), the control point marginalize the result note of constraint do (X ' _Next, Y ' _Next),

$\left\{\begin{array}{c}{X^{\&prime;}}_{\mathrm{next}}=X_{\mathrm{next}}+\mathrm{UnitLength}*\mathrm{\&alpha;}*\sin \mathrm{\&beta;}\\ {Y^{\&prime;}}_{\mathrm{next}}=Y_{\mathrm{next}}+\mathrm{UnitLength}*\mathrm{\&alpha;}*\cos \mathrm{\&beta;}\end{array}\right.$

Wherein:

Sin β represents the component of control point in the x direction;

Cos β represents the component of control point in the y direction.

As above example is supposed

$\left\{\begin{array}{c}{X^{\&prime;}}_{\mathrm{next}}=320+100*2^{\frac{1}{2}}*\frac{1}{\sqrt{5}}=383\\ {Y^{\&prime;}}_{\mathrm{next}}=350+100*2^{\frac{1}{2}}*\frac{2}{\sqrt{5}}=426\end{array}\right.$

7) with Long-distance Control point according to marginalize the constraint the result (X ' _Next, Y ' _Next) be plotted on user's the screen;

8) wipe the control point of previous prediction, draw the control point of up-to-date prediction, the rest may be inferred, and complete in the control point of all predictions in dither process.

Claims (1)

1. control point constrained procedure that adopts marginal effect is characterized in that:

1) the Long-distance Control point of the nearest N of record within second is in the shift position of each millisecond;

2) when network jitter takes place, for the shift position of PREDICTIVE CONTROL point, at first calculation control point is used aveVelocity at the present speed of directions X and Y direction _XAnd aveVelocity _YRepresent; Present speed is by from current nearest T _LastPosition coordinates (X constantly _Last, Y _Last) and the preceding T of M millisecond _FlagPosition coordinates (the X that constantly receives _Flag, Y _Flag) calculate;

$\left\{\begin{array}{c}\mathrm{ave}{\mathrm{Velocity}}_X=(X_{\mathrm{last}}-X_{\mathrm{flag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\\ \mathrm{ave}{\mathrm{Velocity}}_Y=(Y_{\mathrm{last}}-Y_{\mathrm{flag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\end{array}\right.$

3) calculation control point is used aveAcceleration in the current acceleration of directions X and Y direction _XAnd aveAcceleration _YRepresent; Current acceleration is by from current nearest T _Last(X constantly _Last, Y _Last) speed and the T of position _Flag(X constantly _Flag, Y _Flag) another speed of position calculates;

$\left\{\begin{array}{c}\mathrm{ave}{\mathrm{Acceleration}}_X=({\mathrm{Velocity}}_{\mathrm{Xlast}}-{\mathrm{Velocity}}_{\mathrm{Xflag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\\ {\mathrm{aveAcceleration}}_Y=({\mathrm{Velocity}}_{\mathrm{Ylast}}-{\mathrm{Velocity}}_{\mathrm{Yflag}})/(T_{\mathrm{last}}-T_{\mathrm{flag}})\end{array}\right.$

4) calculate the initial predicted position

The current location note at control point is (X _Current, Y _Current), and the initial predicted position at control point note is (X _Next, Y _Next), when network jitter takes place, (X _Next, Y _Next) prediction calculate by following formula:

$\left\{\begin{array}{c}{X}_{\mathrm{next}}=X_{\mathrm{current}}+\mathrm{ave}{\mathrm{Velocity}}_X*T+\mathrm{ave}{\mathrm{Acceleration}}_X*T^2\\ Y_{\mathrm{next}}=Y_{\mathrm{current}}+{\mathrm{aveVelocity}}_Y*T+{\mathrm{aveAcceleration}}_Y*T^2\end{array}\right.$

Wherein:

T is current shake time-delay;

5) calculate the prediction weight

Initially drag the position at control point and the distance between the initial predicted position according to the user, calculate the prediction weight; Distance is far away more, retrains obvious more;

$\mathrm{\&alpha;}=\left\{\begin{array}{cc}1& ,\mathrm{Length}\&le;\mathrm{UnitLength}\\ {(\mathrm{Length}/\mathrm{UnitLength})}^b,0<b<1& ,\mathrm{Length}>\mathrm{UnitLength}\end{array}\right.$

Wherein:

Length is that the user initially drags the position at control point and the distance between the initial predicted position;

UnitLength can decide according to different systems and application;

B is greater than 0 constant less than 1;

The position that initially drags the control point with the user is the center of circle, and UnitLength is that radius forms a scope, if the position of initial predicted in this scope, system keeps the position of original prediction not change, i.e. α=1; When the position of initial prediction exceeds this scope, system adopts (Length/UnitLength) ^bAs weight;

6) carry out the constraint that marginalizes that predict at the control point

Utilize the prediction weight to the constraint that further marginalizes of initial predicted position, the initial predicted position is (X _Next, Y _Next), the control point marginalize the result note of constraint do (X ' _Next, Y ' _Next),

$\left\{\begin{array}{c}{X^{\&prime;}}_{\mathrm{next}}=X_{\mathrm{next}}+\mathrm{UnitLength}*\mathrm{\&alpha;}*\sin \mathrm{\&beta;}\\ {Y^{\&prime;}}_{\mathrm{next}}=Y_{\mathrm{next}}+\mathrm{UnitLength}*\mathrm{\&alpha;}*\cos \mathrm{\&beta;}\end{array}\right.$

Wherein:

β is the described present speed at control point and the angle of Y direction;

7) with Long-distance Control point according to marginalize the constraint the result (X ' _Next, Y ' _Next) be plotted on user's the screen;

8) wipe the control point of previous prediction, draw the control point of up-to-date prediction, the rest may be inferred, and complete in the control point of all predictions in dither process.