In AI, at times, it is a challenging situation to understand the problem itself rather than to find a solution to it. When finding solutions, we tend to go towards two aspects that is the datasets to be used and the AI models. But we tend to ignore another important aspect of the problem which AI environment. The surroundings in which the agent has to work is known as its environment. There are a number of AI Environment types available for different AI Applications.
Basics -AI Environment Types
The agent acts upon the environment with the help of actuators and it uses its sensors to gather information from the environment. The various characteristics of the environment in which the agent has to work must be considered promptly while determining the correct type of agent for the AI system.
There are a number of different types of environments in which the agents work. These may be characterised by the nature of problem, the shape and frequency of data, the amount of knowledge and so on. Hence, a list of some of the types of environments in AI is as follows:
Different AI Environment Types
Single Agent verses Multi Agent AI Environments
This is one of the easiest ways to classify the agent’s environment. This is done on the basis of the number of agents involved in solving the AI problem. When only one single agent is involved in solving the problem, it is referred to as a Single Agent AI environment.
This is an environment in which a majority of AI models work now days. Whereas, when multiple agents work in close collaboration or in competition with each other it is referred to as a Multi Agent AI environment.
Complete verses Incomplete AI environments
Complete AI Environment is the environment which is enough for a problem to be solved completely. Whereas if the AI system is unable to anticipate all the moves in advance in order to solve the problem, it refers to Incomplete AI Environment. Here good equilibrium principles like Nash equilibrium are used.
For instance, Chess is an example of a complete AI environment whereas Poker is an example of Incomplete AI Environment.
Fully Observable verses Partially Observable AI environments
In fully observable AI environments, all the required information in order to complete a task successfully is completely accessible by the AI system. That is the sensor of the agent is able to access the complete state of the agent at a particular point of time to reach at a solution. The partially observable AI environments solve the tasks with a handful of information. These systems mostly rely on the statistical techniques to induce the knowledge from the environment.
For instance, image recognition can operate only if full information about the task is available. Hence it comes under fully observable AI environments. Whereas automated cars are an example of partially observable AI environments.
Competitive verses Collaborative AI environments
In competitive AI environments various AI agents are made to work against each other with an intent to maximize the outcome. This means that the agents compete with each other to move towards the goal. Whereas, in collaborative AI environments different AI agents work together in order to achieve a goal.
For example, games like chess are an example of competitive AI environments as the agents compete against each other in order to win the game of chess. Whereas, automated cars, or smart home sensors work in collaborative AI environments as they work with each other in order to avoid collisions and reach their desired destinations.
Static verses Dynamic AI Environments
As the name suggests, Static AI environments work on the basis of knowledge sources which do not change frequently over a period of time. Static environments form an idle environment for most agents. Whereas, when the data source keeps on changing frequently over a period of time it is referred as dynamic AI environment. Agents in dynamic environment require regular training to adapt to the environment.
For example, speech analysis or empty houses are examples of static AI environments. Whereas, the visuals captured by a drone system changes frequently hence they are an example of dynamic AI environment.
Discrete verses Continuous AI Environments
Discrete AI environments are the ones where a definite set of possibilities can lead the agent towards the required outcome or goal. In continuous AI environments the systems depend on the fast changing, unknown data sources.
For instance, in chess a set of definite moves can lead the system towards its goal. The number of moves made might change but still they are finite in number. So, chess is an example of a discrete AI environment. Whereas, drone systems, automated cars and multi-player video games form examples of continuous AI environments as the environment where they are employed keep on changing rapidly.
Deterministic verses Stochastic AI environments
Deterministic AI environments are the ones where the outcomes can be determined as these environments ignore all the uncertainties. Here, the uniqueness of the current state of the agent determines the next state. In stochastic AI environments the uncertainties in the environment will lead to solutions. Hence, we can say that stochastic AI environments are random in nature so the agent is unable to determine them beforehand.
For instance, most of the AI problems do not work in deterministic situations. They are all classified as stochastic. Take for instance the automated cars. These cars work in uncertain environments say road, traffic etc. and they lead towards a solution with prompt responses.
Episodic and Sequential AI Environment
In Episodic AI environment a series of one-shot actions are required. These actions are taken on the basis of current percept of the environment only. Whereas, in Sequential AI environment an agent works on the basis of past experiences in order to determine the next best action to be taken in order to accomplish the goal.