TWI668043B - Method and system for predicting an object location, and game decision method - Google Patents

Abstract

A method for predicting a falling point includes inputting an environmental state and an input parameter; calculating a first moving distance of the controlled object by using an environmental object estimation model according to the input parameter and the environmental state; and detecting whether the controlled object is within the first moving distance The first environmental object or the second environmental object collides; when a collision occurs, the post-collision path of the controlled object is estimated, and the environmental state is updated, and the second moving distance of the controlled object is calculated according to the post-collision path re-utilization environmental object estimation model. And estimating the stop position of the controlled object when no collision occurs within the first moving distance.

Description

Falling point prediction method, drop point prediction system and game decision Policy method

This case is about a method of landing point prediction, a point-of-sale prediction system and a game decision-making method, and in particular, a method, a system and a game decision-making method for predicting the position of a non-player character in a game.

In recent years, with the rapid development of mobile devices, online instant mobile games have become more and more popular among players. By instantly pairing players into a double or multiplayer battle mode, players can enjoy the fun of playing against real people. However, if there are too few online users in a certain period of time, it will lead to the situation that the player can't find a match after the player goes online. Therefore, in order to let the players enjoy the game at any time, the game developer applies artificial intelligence to the game and lets the computer You can simulate the behavior of real people and play against players.

But artificial intelligence usually requires complicated design and calculation, so if you want to get accurate results, it often takes a lot of time, so you can't give feedback in real-time game, and because of the huge amount of computing, it is for the processor of mobile phone. Words are also a huge load. So how do you The faster calculation without affecting the accuracy, and the effect of real-time feedback is a problem to be solved in the field.

The main object of the present invention is to provide a method for predicting a drop point, a system for determining a drop point, and a method for determining a game, which are mainly to improve the problem that the artificial intelligence is too long to calculate due to the large operation, and utilize the non-player character in the game ( The behavior of Non-Player Character (NPC) is divided into a number of small behaviors that can be estimated, and then multiple small individual trainings are used as training models, so that the predicted NPC behavior can be more accurate and achieve the effect of instant feedback.

In order to achieve the above object, the first aspect of the present invention is to provide a method for predicting a drop point, the method comprising the steps of: inputting an environmental state and inputting a parameter, the environmental state comprising a location of the first environmental object or the second environmental object; The parameter and the environmental state use the environmental object estimation model to calculate a first moving distance of the controlled object; and detect whether the controlled object collides with the first environmental object or the second environmental object within the first moving distance; when a collision occurs, Estimating the post-collision path of the controlled object, and updating the environmental state, calculating the second moving distance of the controlled object according to the post-collision path re-utilization environmental object estimation model; and estimating the control when no collision occurs within the first moving distance The stop position of the object.

A second aspect of the present invention is to provide a drop prediction system comprising: a processor and a storage device. The storage device is electrically connected to the processor for storing an environmental state and an input parameter, where the environmental state includes a location of the first environmental object or the second environmental object. Wherein the processor comprises: an environment An estimation module, a collision detection module, a collision calculation module, and a position calculation module. The environmental object estimation module calculates the first moving distance of the controlled object based on the input parameters and the environmental state. The collision detection module is electrically connected to the environmental object estimation module to detect whether the controlled object collides with the first environmental object or the second environmental object within the first moving distance. The collision calculation module and the collision detection module are electrically connected to calculate a post-collision path of the controlled object after the collision, and update the environmental state, and then calculate the controlled object according to the re-use of the environmental object estimation model according to the post-collision path. The second moving distance. The position calculation module is electrically connected to the collision detection module, and is used to estimate a stop position of the controlled object when there is no collision within the first moving distance.

The third aspect of the present invention is to provide a game decision method, which comprises: obtaining an environmental state; determining a plurality of input parameters; and sequentially applying the environmental state and the plurality of input parameters to the drop prediction method to obtain the controlled object. a plurality of stop positions, the stop positions respectively corresponding to one of the input parameters; estimating the respective values of the stop positions; and determining the optimal input parameters from the input parameters according to the respective values of the estimated stop positions.

The drop point prediction method, the drop point prediction system and the game decision method of the present invention mainly improve the problem that the artificial intelligence has a long calculation time due to the huge operation, and divides the behavior of the NPC in the game into a plurality of small behaviors that can be estimated. Then, multiple small individuals are trained as training models, so that the predicted NPC behavior can be more accurate and achieve the effect of instant feedback.

100‧‧‧Droping Prediction System

101‧‧‧ processor

102‧‧‧Storage device

110‧‧‧Environmental Object Estimation Module

120‧‧‧ collision detection module

130‧‧‧Collision calculation module

140‧‧‧Location Computing Module

200‧‧‧Drop prediction method

400‧‧‧ Game decision method

510‧‧‧ clubs

520‧‧ ‧ cue ball

530a, 530b, 530c‧‧‧a ball

S210~S250, S241~S346, S410~S450‧‧‧ steps

The above and other objects, features and advantages of the present invention are The embodiment can be more clearly understood, and the description of the drawings is as follows: FIG. 1 is a schematic diagram of a landing point prediction system according to some embodiments of the present invention; FIG. 2 is a drawing according to some embodiments of the present invention. A flowchart of a method for predicting a placement point; a third diagram is a flowchart of step S240 according to some embodiments of the present disclosure; and a fourth diagram is a game decision method according to some embodiments of the present disclosure. FIG. 5A is a schematic diagram of a billiard game according to some embodiments of the present invention; and FIG. 5B is a schematic diagram of a billiard game according to some embodiments of the present invention.

The following disclosure provides many different embodiments or illustrations for implementing different features of the invention. The elements and configurations of the specific illustrations are used in the following discussion to simplify the disclosure. Any examples discussed are for illustrative purposes only and are not intended to limit the scope and meaning of the invention or its examples. In addition, the present disclosure may repeatedly recite numerical symbols and/or letters in different examples, which are for simplicity and elaboration, and do not specify the relationship between the various embodiments and/or configurations in the following discussion.

Please refer to Figure 1. 1 is a schematic diagram of a drop prediction system 100, in accordance with some embodiments of the present disclosure. As depicted in Figure 1 The drop prediction system 100 includes a processor 101 and a storage device 102. The storage device 102 is electrically connected to the processor 101 and configured to store an environmental state and an input parameter, where the environmental state includes a location of the first environmental object or the second environmental object. The processor 110 further includes an environment object estimation module 110, a collision detection module 120, a collision calculation module 130, and a position calculation module 140. The environment object estimation module 110 is electrically connected to the collision detection module 120. The collision detection module 120 is electrically connected to the collision calculation module 130 and the position calculation module 140.

In various embodiments of the present invention, the processor 101 can be implemented as an integrated circuit such as a micro control unit, a microprocessor, a digital signal processor, or an application specific integrated circuit (application specific) Integrated circuit (ASIC), logic circuit or other similar component or combination of the above. The storage device 102 can be implemented as a memory, a hard disk, a flash drive, a memory card, or the like.

Please refer to Figure 2. 2 is a flow chart of a method of landing prediction method 200, according to some embodiments of the present disclosure. The placement prediction method 200 of an embodiment of the present invention can be used to calculate NPC behavior. The NPC behavior referred to herein can refer to the physical state of all non-home characters in the game, for example, NPC behavior in a pool game. It can refer to the relationship between the cue ball, the sub-ball and the table, and the physical state of the cue ball and the sub-ball. In an embodiment, the placement prediction method 200 shown in FIG. 2 can be applied to the placement prediction system 100 of FIG. 1 , and the processing unit 110 is configured to perform NPC according to the steps described in the following placement prediction method 200 . Estimation of behavior. As shown in FIG. 2, the placement prediction method 200 includes the following steps: Step S210: input an environmental state and input parameters; Step S220: Calculate a first moving distance of the controlled object by using an environmental object estimation model according to the input parameter; Step S230: Detect whether the controlled object is within the first moving distance and the first Collision of the environmental object or the second environmental object; step S240: estimating a post-collision path of the controlled object and updating the environmental state if a collision occurs; and step S250: estimating the controlled object if no collision occurs within the first moving distance Stop position.

In step S210, an environmental state and an input parameter are input, and the environmental state includes a location of the first environmental object or the second environmental object. In an embodiment, the first environmental object may be a sub-ball, and the second environmental object may be a table. The input parameters may be a force of hitting the ball, a position of the hitting ball, and a direction of hitting the ball. The invention only needs to change the environment state of the input and the input parameters to be applicable to different instant competition games (for example, baseball games, game play, etc.), and is not limited to use in the pool game of instant competition.

Before performing step S220, the motion state of the cue ball, the physical state of the cue ball and the collision of the sub-ball, and the physical state of the collision between the cue ball and the table are separately divided into a plurality of models for training, and the plurality of models are respectively the environmental object estimation model. The pre-collision velocity model and the tangent offset model, and the environmental object estimation model includes three sub-models, namely the base distance sub-model, the state switching sub-model and the rolling speed attenuation sub-model.

First, the base distance submodel is introduced. When the cue ball has stopped without hitting anything (a ball or a table), the cue ball is stressed in this embodiment. After the stop until there is no collision between the movements, as the base distance for the cue ball movement. Enter the distance that the cue ball moves under different ball striking forces, and then use the neural network or Linear Regression to generate the base distance submodel, so that you can directly input the battering force path through the base. The distance from the submodel outputs the base distance at which the cue moves.

Next, the pre-collision velocity model is introduced. Before the collision calculation, the velocity before the cue ball collision must be known. Therefore, the initial velocity of the cue ball under the corresponding ball striking force, the initial angular velocity, and the movement of the cue ball to the collision point must be calculated. Distance and other parameters. First, enter the initial velocity under different ball striking force, the initial angular velocity, the moving distance from the cue ball to the collision point, the velocity before the collision, and the angular velocity before the collision. Then use the neural network or linear regression to generate the collision. The velocity model allows the initial velocity, the initial angular velocity, and the moving distance of the cue ball to the collision point to be directly input, and the velocity and angular velocity of the cue ball before the collision are output by the pre-collision velocity model.

Next, the tangent offset model is introduced. When the cue ball and the sub-ball collide, an almost complete elastic collision occurs (the cue ball moves in the tangential direction after the collision), so according to the pre-collision speed, the pre-collision angular velocity, and the cue ball and the sub-ball. The angle between the balls can be used to infer the speed and angular velocity of the cue ball and the sub-ball collision. However, in addition to the kinetic energy on the cue ball, it will be transferred to the sub-ball. Because of the frictional force, the moving direction of the cue ball and the original tangential direction will produce all the line offset angles. The angular velocity and the angle between the cue ball and the subsphere are used to construct a tangent offset model to obtain the tangent offset angle. Of course, the cue ball also has a ball condition that collides with the table. If it hits the table, it is profitable. The traveling direction of the cue ball is calculated by the principle that the incident angle is equal to the reflection angle, and then the direction after the collision is corrected according to the pre-collision speed and the pre-collision angular velocity.

Then introduce the rolling speed attenuation sub-model. In order to simulate the real situation of the cue ball and the sub-ball rolling, it is necessary to generalize the kinetic energy loss of the ball and the table friction during rolling to a function of attenuation, because the ball will go through multiple sliding or rolling states. Switching, the more times the state is switched, the more the speed of the ball will decay. Therefore, it is necessary to calculate how many times the ball is switched during the movement, and the number of times is substituted into the attenuation function to estimate the attenuation speed of the ball.

Then introduce the state switching submodel. After observing the motion state of the cue ball, it can be found that after the cue ball is hit or the cue ball passes the collision, the cue ball will slide for a certain distance and then roll for a distance, and the speed of the sliding state is attenuated and scrolled. The velocity decay of the state is different, the velocity decay of the sliding state is attenuated linearly, and the velocity decay of the rolling state is exponentially attenuated. Therefore, after the moving distance of the sliding state ball can be calculated, the moving distance of the rolling state ball can be inferred by subtracting the moving distance of the mother ball to the collision point and the moving distance of the sliding state ball.

After the base distance sub-model, the state switching sub-model, the rolling speed attenuation sub-model, the pre-crash speed model, and the tangent offset model are trained, the estimation of the cue ball and the sub-ball movement path can be started.

Referring to FIG. 2 again, in step S220, the first moving distance of the controlled object is calculated according to the input parameter using the environmental object estimation model. In an embodiment, the distance of the cue ball before the collision can be calculated by using the base distance sub-model, the state switching sub-model, and the rolling speed attenuation sub-model.

In step S230, it is detected whether the controlled object collides with the first environmental object or the second environmental object within the first moving distance. In this embodiment, whether the previously calculated distance of the cue ball before the collision and the motion vector are used to determine whether the vector of the other environmental objects (the sub-ball and the table) overlap in the motion vector.

In step S240, if a collision occurs, the post-collision path of the controlled object is estimated, and the environmental state is updated. For the detailed process of step S240, please refer to FIG. 3, which is a flowchart of step S240 according to some embodiments of the present disclosure. As shown in FIG. 3, step S240 includes the following steps: Step S241: If a collision occurs, the velocity before the collision of the controlled object is calculated by using the pre-collision velocity model; and step S242: determining whether the controlled object collides with the first environmental object. Step S243: If a collision event occurs between the controlled object and the first environmental object, the tangential offset model is used to calculate the angular offset after the collision between the controlled object and the first environmental object; and step S244: calculating the first collision of the controlled object a post-speed and a second post-collision speed of the first environmental object; step S245: calculating the controlled object and the second environment if the controlled object does not collide with the first environmental object but collides with the second environmental object The angular offset after the collision of the object and the speed after the collision; and step S246: recording the occurrence position of the collision event, and updating the environmental state according to the occurrence position, the first collision speed, and the second collision speed.

In step S241 and step S242, if a collision occurs, the velocity before the collision of the controlled object is calculated by using the pre-collision velocity model, and then it is determined whether the controlled object collides with the first environmental object. After step S230 determines whether a collision has occurred, if a collision occurs, it is necessary to first calculate the remaining speed and angular velocity of the pre-collision cue ball using the previously established pre-crash velocity model, and then further determine that the cue ball is the first environmental object or the second. Collision of environmental objects (sub-balls or tables).

In step S243 and step S244, if a collision event occurs between the controlled object and the first environmental object, the tangential offset model is used to calculate an angular offset of the controlled object after colliding with the first environmental object; and then the first object of the controlled object is calculated. The velocity after a collision and the calculation of the second post-collision speed of the first environmental object. In an embodiment, if the cue ball collides with the sub-ball, the cue ball and the sub-ball are first preset as elastic collisions, the angle between the cue ball and the sub-ball collision is calculated, and the previously obtained cue ball velocity before the collision is obtained. Compared with the angular velocity and the angle of the cue ball after the elastic collision calculation, the angle of the cue ball is substituted into the tangent offset model to obtain the corrected post-collision angle, and then calculate the velocity and angular velocity of the cue ball and the sub-ball collision respectively. The angle after the collision of the sub-ball is the angle calculated by the elastic collision.

In step S245, if the controlled object does not collide with the first environmental object, but collides with the second environmental object, the angular offset and the post-collision speed after the collision between the controlled object and the second environmental object are calculated. In reality, depending on the impact of the cue ball on the table, the degree of sinking of the table, the angular velocity of the cue ball, and the strength of the friction will have a large image of the bounce angle behind the ball and the table. In order to simplify the calculation, in this embodiment, if the mother The ball collides with the table, and the impact surface of the cue ball and the table is regarded as a plane perpendicular to the table top. Therefore, the angle after the cue ball hits the table can be calculated theoretically by the angle of incidence equal to the angle of reflection.

In step S246, the occurrence position of the collision event is recorded, and the environmental state is updated according to the occurrence position, the first post-collision speed, and the second post-collision speed. In an embodiment, the position where the cue ball collides with the sub-ball or the cue ball hits the table is recorded, and then the sub-ball is removed from the original position, and the collision position and the impact of the cue ball and the sub-ball are The speed and angular velocity are passed back to the list of recorded environmental conditions and become the parameters for the next cue ball and sub-ball movement. The position and moving distance of the cue ball and the sub-ball after the collision will be recalculated from step S220, so when the cue ball is knocked out, unless it is stopped without hitting the sub-ball or the table, the other situation is cue ball. Or the ball will go through the calculation of the multiple point prediction method 200 until it stops, so there will be many collision phases.

In step S250, if no collision occurs within the first moving distance, the stop position of the controlled object is estimated. In an embodiment, if the cue ball does not hit the sub-ball or the table after being hit, the cue ball will gradually stop due to friction with the table top. After calculating the stop position, the cue position is updated and stored in the record. In the list of environmental states. If the cue ball stops without an impact event, it means that the movement of the cue ball ends. If the force of the hitting ball is not input, the movement of the cue ball will not start.

In another embodiment, the motion of the sub-ball struck by the cue ball can also be estimated by using the drop prediction method 200, and the velocity and angular velocity generated by the impact of the sub-ball as an input parameter, by judging the impacted sub-ball Whether there is a collision with the table or other sub-balls, repeat the steps S220~S240 Count, until the impacted ball stops.

Please refer to Figure 4. 4 is a flow diagram of a game decision method 400, illustrated in accordance with some embodiments of the present disclosure. The game decision method 400 of an embodiment of the present invention can be used to automatically select an optimal result from a plurality of possibilities such that a better result is achieved in combating the player. As shown in FIG. 4, the game decision method 400 includes the following steps: Step S410: Acquiring an environmental state; Step S420: Determining a plurality of input parameters; Step S430: Applying the environmental state and the input parameters to the aforementioned Falling Point Prediction method in sequence Obtaining a plurality of stop positions of the controlled object, the stop positions respectively corresponding to one of the input parameters; step S440: estimating the respective values of the stop positions; and step S450: according to the respective values of the estimated stop positions, by the input parameters Determine the best input parameters.

In steps S410 and S420, an environmental state is acquired, and then a plurality of input parameters are determined. In one embodiment, the environmental state includes the position of the sub-ball and the cue ball, and in the pool game, it is possible to select how much force to use to hit the cue ball, the table or a certain position, and thus has the possibility of multiple hitting.

In step S430, the environmental state and the input parameters are sequentially applied to the foregoing landing point prediction method 200, and a plurality of stop positions of the controlled object are obtained, and the stop positions respectively correspond to one of the input parameters. Please refer to Figure 5A and Figure 5B together. FIG. 5A is a schematic diagram of a billiard game according to some embodiments of the present invention, and FIG. 5B is a schematic diagram of a billiard game according to some embodiments of the present invention. As shown in Figure 5A, in this embodiment In the middle, the cue ball 520 has three hitting options, respectively, which can impact the sub-balls 530a, 530b, and 530c, and the position of the cue ball 520 after hitting the sub-balls 530a, 530b, and 530c is calculated by the drop prediction method 200, so three calculations are calculated. Different cue ball placement locations.

In the above, in steps S440 and S450, the respective values of the stop positions are estimated, and the optimal input parameters are determined from the input parameters based on the respective values of the estimated stop positions. In the embodiment of FIGS. 5A and 5B, the cue ball 520 strikes the sub-ball 530a such that after the sub-ball 530a enters the hole, the cue ball 530b can be subsequently struck if the cue ball 520 is parked in place. If the cue ball 520 first hits the sub-ball 530b or the sub-ball 530c, since the distance between the cue ball 520 balls 530b and 530c is far, the accuracy of controlling the stop position of the cue ball 520 is not good, resulting in the final stop of the cue ball 520. The position is not good enough to play a ball, so the distance between the cue ball 520 and the sub-balls 530a, 530b and 530c can be used as the standard for selecting the ball. Of course, the invention is not limited thereto. The rules of the game of pool are used as a criterion. Therefore, the club 510 finally moves to the position where the ball 530a is to be hit, and the ball 520 can be caused to hit the ball 530a, and the ball 530a is put into the bag and the ball 520 is stopped in place to continue the ball 530b. .

It can be seen from the above-mentioned embodiments of the present invention that the problem of excessive computational time caused by the huge computational complexity of the prior art is improved, and the physical scrolling behavior of the mother ball and the sub-ball in the game is divided into a plurality of small behaviors that can be estimated, and then A plurality of small-sized individual trainings are used as training models, so that the physical rolling behavior of the predicted cue ball and the sub-ball can be more accurate and achieve the effect of instant feedback.

In addition, the above examples include sequential steps, but these The steps need not be performed in the order shown. Performing these steps in a different order is within the scope of the present disclosure. Such steps may be added, substituted, altered, and/or omitted as appropriate within the spirit and scope of the embodiments of the present disclosure.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make various changes and refinements without departing from the spirit and scope of the present case. The scope defined in the patent application is subject to change.

Claims (11)

A method for predicting a drop point includes: inputting an environmental state and an input parameter, the environmental state comprising a location of a first environmental object or a second environmental object; calculating an environment object estimation model according to the input parameter and the environmental state a first moving distance of the controlled object; detecting whether the controlled object collides with the first environmental object or the second environmental object within the first moving distance; and when the collision occurs, estimating the controlled object a post-collision path, and updating the environmental state by using a collision position, and reusing the environmental object estimation model according to the post-collision path to calculate a second moving distance of the controlled object; and when within the first moving distance No collision occurred, estimating the stop position of one of the controlled objects. The method for predicting a drop point according to claim 1, further comprising: if a collision occurs, calculating a speed before the collision of the controlled object by using a pre-crash speed model, and then determining that the controlled object is associated with the first environmental object and A collision event occurs in one of the second environmental objects. The method for predicting a drop point according to claim 2, further comprising: if the collision event occurs between the controlled object and the first environmental object, using the line offset model to calculate the collision between the controlled object and the first environmental object An angular offset; and Calculating a first post-collision speed of the controlled object and calculating a second post-collision speed of the first ambient object. The method for predicting a drop point according to claim 2, further comprising: if the collision event occurs between the controlled object and the second environmental object, calculating the angular offset after the controlled object collides with the second environmental object And the speed after a collision. The method for predicting a drop point according to claim 3, further comprising: recording the collision position of the collision event, and updating the environmental state according to the collision position, the first post-collision speed, and the second post-collision speed. A drop point prediction system includes: a processor; a storage device electrically connected to the processor, configured to store an environmental state and an input parameter, the environmental state comprising a first environmental object or a second environmental object The processor includes: an environmental object estimation module, and calculating a first moving distance of the controlled object according to the input parameter and the environmental state; a collision detecting module, and the environment object Estimating a module electrical connection for detecting whether the controlled object collides with the first environmental object or the second environmental object within the first moving distance; a collision calculation module, and the collision detection mode Group electrical connection Calculating a post-collision path of the controlled object after the collision, and updating the environmental state by using a collision position, and then reusing the environmental object estimation model according to the post-collision path to calculate a second moving distance of the controlled object; And a position calculation module electrically connected to the collision detection module for estimating a stop position of the controlled object when no collision occurs within the first movement distance. The landing point prediction system of claim 6, wherein the collision calculation module is further configured to calculate a speed before the collision of the controlled object by using a pre-crash speed model, and then determine that the controlled object is related to the first environment. A collision event occurs in one of the object and the second environmental object. The landing point prediction system of claim 7, wherein the collision calculation module calculates the controlled object and the first environment by using a line offset model if the collision event occurs between the controlled object and the first environment object An angular offset after the object collides, and then calculating a first post-collision speed of the controlled object and calculating a second post-collision speed of the first ambient object. The landing point prediction system of claim 7, wherein if the collision event occurs between the controlled object and the second environmental object, the collision calculation module calculates the collision between the controlled object and the second environmental object. Angle offset and a post-collision speed. The landing point prediction system of claim 7, wherein the location calculation module records the collision location of the collision event, and updates the environment according to the collision location, the first post-collision speed, and the second post-collision speed. status. A game decision method includes: obtaining an environmental state; determining a plurality of input parameters; applying the environmental state and the input parameters to the drop prediction method as described in item 1 to obtain a plurality of the controlled objects Stop positions respectively corresponding to one of the input parameters; estimating a value of each of the stop positions; and determining, according to the estimated respective values of the stop positions, determined by the input parameters An optimal input parameter.