Study Notes on game AI

Review notes
Chapter 1
AI is a branch of AI science and technology.
Artificial Intelligence games are designed to apply Theories, Methods, and technologies of artificial intelligence to the development of intelligent games and improve the intelligence level of game roles, it mainly includes the theoretical methods and technologies of intelligent game role action, perception, response, emotion, memory, and learning.
A role also acts as a role. It refers to the act of actors performing fictitious characters based on scripts in the form of performance arts, and also to the act of fictitious characters.
The player role (PC) controller is like the role's brain. The input is the status information of the game world, and the output is the action choice that affects the game world and leads to the corresponding behavior of non-player roles. A game role controlled by a player using a joystick or other input devices.
Non-player role (NPC): Other roles
Role Behavior Control: each role has at least one controller associated with it, and the controller can be shared among different roles.
The Controller is like the role's brain. The input is the status information of the game world, and the output is the action choice that affects the game world and leads to the corresponding behavior of non-player roles.
Player role Controller: explains the various operations of the player's joystick and the player's brain.
Non-player role controller (NPC): It has multiple forms and different functions.
Game System Structure
Game status: Describes all objects in the game world and their status information.
Simulator: "game physics" to develop rules on how game states change
Renderer: uses the geometric relationships and textures of the objects in the game to provide game state profiling.
Controller: Responsible for role selection
Capabilities of non-player roles: Action and perception
Non-player role controller design method: directly follows a set of pre-prepared rules of action, automatic search, and learning capabilities
Artificial Intelligence: intelligence is a comprehensive ability to understand objective things and use knowledge to solve problems.
Capabilities of intelligence: perception, memory and thinking, learning and adaptability, and Behavior
From the perspective of capabilities: artificial intelligence is relative to human natural intelligence. The so-called artificial intelligence refers to the intelligence achieved on machines using artificial methods.
From the perspective of the subject, this is a discipline that studies how to construct intelligent machines or intelligent systems so that they can simulate, extend, and expand human intelligence.
As for how to measure whether a machine is smart, the Turing test.
-- Chapter 1 end -----
Chapter 2

Knowledge: Knowledge is the foundation of all smart behaviors and an important object of AI research. The intelligence of an intelligent system depends largely on the quantity of knowledge and the degree to which it can be used. Knowledge is also the recognition and experience accumulated by people in the practice of transforming the objective world.

Description of the objective world includes data (a combination of symbols or symbols) information (a structure composed of different data) knowledge (understanding of the regularity of the objective world formed by Intelligent Processing of Information)

Knowledge Representation of State Space Law: uses the test search method to find a solution in a possible solution space to solve the problem. This is a question representation and Solution Method Based on the answer space. It represents and solves the problem based on the state and operator.

State Space Method: from an initial state, an operator is added each time, and the test sequence of the operator is established progressively until the target State is reached.

Status chart: a chart of nodes corresponding to each State where the initial state can be reached. Finding an operator sequence problem that transforms from one State to another is equivalent to finding a path problem in a graph.

The following concepts need to be mastered: graphs, Directed Graphs, successor nodes, parent nodes, paths, and costs
State: an ordered set of minimum variables introduced to describe the differences between different things.
State variable: each element in the state set
Specific Status: A group of values for each component
Operator: it is also called an operator to convert a problem from one state to another. Operators can be steps, procedures, rules, mathematical operators, operator numbers, or logical symbols.
Status space of the problem: shows the diagram of all possible states and their relationships of the problem. It contains three sets of description, namely, the initial state set of all possible problems, operator set F and target State set g state space can be recorded as triple (S, f, g)
---- Chapter 2 end ------
Chapter 3
Search: based on the actual situation of the problem, we constantly look for available knowledge to construct a low-cost reasoning route so that the problem can be solved.
Search classification: Classified by whether or not to use heuristic information (blind search and heuristic search)
Divided by the problem representation (status space search, AND/OR Graph Search)
The state space method and the problem reduction method are the two most basic problem solutions in AI. The State Space Representation Method and the and/or tree representation method are the two most basic problem Representation Methods in AI.
General Graph Search Data Structure: Open Table-used to store the generated nodes. Because these nodes have not been extended, Open table can also be called an unextended node table.
Closed table-used to store extended nodes. Therefore, closed tables are also called extended node tables.
------- Miss a lot -------------

Chapter 5
Genetic algorithm: it is a search algorithm based on the information Genetic Mechanism in the evolutionary process and the natural selection principle of survival of the fittest. It uses the probability search process to search in the state space to generate a new sample. The features are universal, robust, sub-optimal, and satisfactory solutions.
Genetic algorithms are different from traditional optimization algorithms:
1. The genetic algorithm does not directly act on the parameter set, but uses some encoding of the parameter set.
2. The genetic algorithm does not start from a single point, but searches for a point in a group.
3. Genetic Algorithms use adaptive value information without derivative or other auxiliary information.
3. Genetic Algorithms use probability transfer rules rather than deterministic rules.
Genetic algorithm preparation:
1. Confirm the Representation Scheme
2. Determine the measurement of the adaptive Value
3. determine the parameters and variables that control the algorithm.
4. Determine how to specify the results and the criteria for program running termination
Steps:
Encoding-initial group generation-fitness evaluation-selection-cross operation-variation-generation of new groups-repetition
Steps for genetic algorithms:
(1) encoding: Data of the space to be parsed is represented as the data of the genotype string structure.
(2) initial group generation: according to the set encoding scheme, n initial String Structure Data are randomly generated to form a population and start iteration as the initial point.
(3) Adaptive evaluation Detection: the fitness function evaluates the scores of each seed, indicating the merits and demerits of the seed.
(4) Selection: selects excellent seeds from the current population as the parent generation to breed the next generation. It is commonly used to implement round-robin.
(5) hybridization: a new generation of individuals can be obtained based on the set hybridization rate. New individuals combine the characteristics of their parent individuals.
(6) variation: changes the value of a string in the string structure data according to the variation rate of the selected individual.
--- Chapter 5 end ---------------