31results about How to "Increase winning percentage" patented technology

The invention discloses an agent learning method based on knowledge guidance-tactical perception. In the environment of a two-party fight game, human knowledge is used to train an agent which fights with a specific tactical strategy through two stages of training. The method comprises the steps that 1) a game screen and the state of the agent are encoded to form a state feature; 2) a script with aspecific fight strategy is constructed by manually writing a script; 3) the agent fights with the script to collect the state and motion data of an opponent to carry out training in the first stage;4) the trained agent fights with a computer built-in AI, and training in the second stage is carried out through reinforcement learning; and 5) the learning framework is used to train the agent to participate in fight with specific tactics. The method provided by the invention is applicable to agent training in a two-party fight mode in a micro-operation environment, and acquires a good winning rate in the face of various micro-operation fight scenes.

The invention discloses a combined type toy spinning top capable of automatically splitting. The combined type toy spinning top is characterized by comprising at least two spinning top bodies which are butted vertically, wherein the spinning top tip of each spinning top body is an elastic spinning top tip; the upper spinning top body in the spinning top can automatically separate from the lower spinning top body when being impacted or blocked during the rotation, and ejects out under the action of the elastic spinning top tip of the upper spinning top body, so that the two spinning top bodies rotating independently respectively are formed. According to the invention, when a user plays the spinning top, the fact that the spinning top can be split into two parts or three parts during a competition can be realized, so that the attack force of the spinning top is greatly improved, and further the win rate is higher; besides, the spinning top tips of the spinning top bodies are elastic spinning top tips, and when the spinning top is split during a game, the upper spinning top body ejects out mainly under the action of the elasticity of the spinning top tip of the upper spinning top body, so that the impact to the lower spinning top body is not large; besides, the soft landing can be realized through the utilization of the buffer action of the elasticity during landing, so that the spinning top tips can be effectively protected, and the operation stability can be kept.

The invention provides a toy gyro good in defensiveness. The toy gyro comprises a spiral cover, a spiral sheet, a spiral base and a hillock point, and is characterized by further comprising an elastic shock absorption piece for buffering impact force when the gyro is impacted. The shock absorption piece is provided with an elastic part used for resisting impact and a connection part fixedly connected with the spiral base. The elastic part is located on the periphery of the spiral base through fixed connection of the connection part and the spiral base, and therefore when the gyro competes with other gyros and the other party impacts on the elastic part, impact force will be reduced due to elastic deformation of the elastic part; thus, the shock absorption effect is achieved to make the gyro rotate stably, anti-impact force is produced and returned to the other party at the same time to interfere the other party, and accordingly the win rate of the gyro is improved. The toy gyro is strong in interestingness, the elastic property of the elastic shock absorption piece can be changed by a player according to judgments so as to achieve a better anti-impact effect, the method to play the toy gyro is novel, the toy gyro can attract more players, and manipulative ability and competitive ability of children can be trained at the same time.

The invention provides a bounceable toy gyro. The toy gyro comprises a gyro cover, a gyro piece, a gyro base and a gyro tip, wherein the gyro cover is located at the top; the gyro piece is located below the gyro cover; the gyro base is used for fixing the gyro piece; and the gyro tip is located at the bottom. The toy gyro is characterized in that a bouncing moving part is arranged between the gyro base and the gyro tip, fixedly connected with the gyro tip and connected with the gyro base in an up-down bouncing manner, so that when the gyro falls to the ground from a launcher, the bouncing moving part and the gyro tip move upwards under the effect of the reactive force of the gravity, then recover elastically, conduct downward returning movement and then move upwards again under the effect of inertia force, the process is repeated, bouncing actions are formed, and the gyro is quite interesting; in addition, in competitions, the gyro can bounce to hit a gyro of an opponent downwards, so that the gyro of the opponent is easier to damage, the aggressiveness of the gyro is greatly increased, and the winning rate is higher; besides, a playing method of the gyro is novel, so that more players like the gyro; and meanwhile, the operational ability and the competitive ability of children can be trained.

The invention discloses an object control method and device, a storage medium and an electronic device. The object control method includes the steps that during execution of a game task by a first target object controlled by a client, a target angle generated by the first target object during an action combination is detected, and the action combination includes two or more target actions executedin the same direction; under the condition that the target angle reaches a trigger threshold, a status adjustment instruction is automatically triggered; in response to the status adjustment instruction, the movement state of the first target object is adjusted from the first state to the second state, and the first time taken by the first target object to complete the game task in the first state is longer than the second time taken by the first target object to complete the game task in the second state; and the first target object is controlled to execute the game task in the second state.The object control method and device, the storage medium and the electronic device solve the technical problem of increasing control time due to high operation complexity.

The present invention discloses an event display method and apparatus. The method comprises the steps of detecting whether a first event executed by an application client in a current cycle is finished, wherein the first event is used for obtaining a first event object meeting a first event result among multiple event objects; and when it is detected that the first event is finished, displaying result information in a predetermined display area of the application client, wherein the result information comprises a first figure identifier for identifying the first event object, and a second figure identifier for identifying a second event object, the second event object is an event object that meets the first event result and is obtained in execution of a second event, and the second event is an event executed by the application client in a previous cycle of the current cycle. The existing display method is low in display efficiency. The event display method solves the problem in the existing display method.

The invention relates to a multi-agent reinforcement learning method and system based on population training. The method comprises the steps of obtaining a first training set according to game videos;training the multi-layer full convolution LSTM network by using the first training set to obtain a first intelligent agent; utilizing the first intelligent agent to perform self-gaming, and obtaininga first population after a set time period; selecting a second intelligent agent, a first intelligent agent set and a second intelligent agent set from the first population; utilizing the first intelligent agent to fight with the selected three groups of intelligent agents at the same time, and storing and updating the first population until any one of the selected three groups of intelligent agents goes wrong, so as to obtain a second population; selecting a replacement agent from the second population to replace the battle agent to continue to fight with the first agent, storing and updating the second population, and obtaining a third population; and until the number of the agents in the third population reaches a preset value, outputting the first agent. According to the invention, the intelligent agent capable of simulating unmanned system combat command and control can be trained.

The invention belongs to the technical field of intelligent decision making, and relates to a war game multi-entity asynchronous collaborative decision making method and device based on reinforcement learning, and the method comprises the steps: obtaining a war game deduction environment and a multi-entity asynchronous collaborative decision making problem, and carrying out the modeling analysis of the multi-entity asynchronous collaborative decision making problem, and obtaining an initial model; according to the initial model, a multi-agent deep reinforcement learning algorithm is adopted to establish an agent network model and a hybrid evaluation network model; training the agent network model and the hybrid evaluation network model to obtain a collaborative decision framework; a loss function of the multi-agent deep reinforcement learning algorithm is reconstructed by setting a weighting operator or optimizing the multi-agent deep reinforcement learning algorithm through multi-step return; using the reconstructed loss function to update the collaborative decision framework; and according to the updated collaborative decision framework, performing decision making on asynchronous collaboration of multiple entities. According to the method, multi-entity asynchronous collaborative decision-making in war game deduction can be realized.

The invention relates to a strategy improvement method for a search game tree on Go. The method comprises the following steps: establishing a search tree by taking a current state as a root node; selecting a child node of the root node for simulation, and if a child node is not simulated, randomly selecting a child from the child node of the root node for simulation; if all the child nodes are simulated for at least one time, selecting the child node with the highest UCB sub-tree; starting to simulate from the selected child node to the leaf node; combining the simulation strategy with uniform sampling and a minimum and maximum strategy; propagating the final simulation result reversely to the root node, and updating the Q value and the N value of the action value function of all leaves on the path; and repeating the steps for many times, and finally selecting the node with the highest utilization item score in the UCB. According to the invention, the improved algorithm is applied to the search strategy of the Go, the evaluation of GNUGo and CGOS is passed, and the final experimental result shows that the algorithm can improve the accuracy of game search in the Go.

The invention provides a threat evaluation method for an air target in an over-the-horizon air combat, and relates to the technical field of threat evaluation in the over-the-horizon air combat. The method comprises the steps of firstly establishing a missile model, and calculating flight data of a missile carried by a target enemy plane; determining a turning angle when the aircraft escapes according to the azimuth angle of the enemy aircraft, and further determining an escape route and a folded escape distance of the aircraft; determining the turning angle and the turning time of the enemy plane according to the azimuth angle of the enemy plane and the speed vector direction of the enemy plane; judging whether the aircraft is within the range of the enemy missile or not, and judging theattack result of the enemy aircraft to the aircraft and the total attack time by combining the flight speeds of the enemy aircraft and the aircraft; and finally, determining a threat evaluation valueof the enemy plane to the aircraft according to an attack result of the enemy plane to the aircraft. According to the method, energy is used as a basis, time is used as a unified dimension, and the judgment precision of absolute threat evaluation is effectively improved; and the output quantity units are unified, so that the judgment precision of relative threat evaluation is effectively improved.

The invention discloses a control system of a soccer robot. The control system comprises a visual processor and a motion controller, wherein the visual processor comprises an image processing module,a positioning module and a communication module, wherein the image processing module is used for processing an image collected by the soccer robot so as to generate image information, the positioningmodule is connected with the image processing module and is used for processing the image information so as to generate the position and distance information of the soccer robot, and the communicationmodule is connected with the positioning module and is used for sending the position and distance information to other soccer robots and receiving instruction information sent by the other soccer robots; and the motion controller includes a motion state management and decision making module, wherein the motion state management and decision making module is connected with the visual processor andis used for receiving the position and distance information and instruction information, and controlling the movement of the soccer robot according to the position and distance information and instruction information. With the control system of the soccer robot of the present invention adopted, the teamwork of a plurality of soccer robots can be realized.

The invention belongs to the technical field of mobile phone MOBA games, and discloses a method, system and device for recommending equipment in real time based on MOBA games. The MOBA games are installed in mobile terminals, at least two mobile terminals perform online games, and the method for recommending equipment is applied to a server. The method comprises the following steps: in response tobattle requests generated by at least two mobile terminals with battle requests, establishing a new game room, and identifying a game role of each mobile terminal; acquiring game attributes of each enemy game role participating in battle at present, performing optimal comparison on the average value of the enemy game attributes, and performing optimal comparison on the basis of the average valueof the enemy game attributes; and recommending equipment purchase based on the screened hostile game attribute average value, and displaying the recommended purchased equipment in the client corresponding to the game role of each friend mobile terminal.

The invention discloses a polypropylene short-term Delta sequence query method. The method quickly queries the short-term Delta sequence by searching the period and serial number. The Delta sequence is an objectively existing sequence that does not depend on human will. The mean value and variance can be used for better trading. When investors know the mean value of the short-term Delta sequence, they can conduct corresponding transactions according to different trading strategies. However, when trading, they should not blindly trade according to time. Wait for the verification signal of the trend change within the variance time, so that the winning rate of the transaction can be improved. Compared with traditional technical analysis, DeIta is more trustworthy and reassuring, because it is the law that exists in the market and guides the direction of the market, while traditional technical analysis is to process the emerging data to make a big difference. A guess that the probability is expected to go the same way is essentially a prediction based on the further continuation of the trend rather than an objective law.

The invention discloses a stock trading method based on reinforcement learning, and relates to the field of machine learning and quantitative trading. According to the invention, stock transaction is carried out based on a cyclic reinforcement learning algorithm with adaptive capability; the method comprises the steps that a user interaction interface and a classification neural network N train a classification neural network; through a circular reinforcement learning RRL training stage, three types of operations of buying, holding and selling are respectively executed in recognized different stock market cycle scenes, after RRL training is completed, an automatic transaction execution and loss stopping stage is entered, dynamic loss stopping is performed according to a loss stopping strategy set by a user, and automatic transaction execution is performed. According to the invention, the specific preference of the user for risk earnings can be satisfied, the risk of manual transaction errors is reduced, and the cost of manual decision making is reduced; compared with a traditional linear model and a Q learning method, the price self-adaptability is higher, the made investment decision is timely and effective, and the transaction winning rate can be greatly improved.