Autonomous Agents
7.1.8 The role of autonomous agents acting within a larger system
IB Subtopic
Definitions:
Agents: anything that can perceive its environment(usually through sensors) and act upon it.
Robotic Agents: Physical machines that use sensors to act upon their environment.
Software Agents: Programs that use information/data to act upon virtual environments.
Autonomous Agents: Agents that use pre-programmed or machine-learned algorithms to perform actions in different situations.
Environments
Agents may have some influence over their environments but many times they can not have complete control over it. Through the use of a control system, the can receive feedback/inputs and act upon it. There are different types of environments:
| Easy to Programm | Hard to Program | |
| Accessibility | Accessible: Can obtain complete/accurate information about the environment | Inaccessible: Can NOT obtain accurate information |
| Uncertainty | Deterministic: Whether we know if the next state of the an environment with certainty. | Non-deterministic: We’re not sure what the state of the the environment will be because of our actions or other effects. |
| Episodic | Episodic: When actions are contained and not affected by previous actions. | Non-episodic: When actions are affected by previous actions. Most real-world environments fall under this definition. |
| Change | Static: The environment doesn’t change. | Dynamic: The environment changes. |
| Continuity | Discrete: The number of states in which a control system can be in at any given time can be guessed and understood ahead of time. | Continuous: Infinite amount of states can be possible from any given state. |
Types of Agents
- Domestic Robots
- Unmanned Vehicles
- Warehouse Robots
- Video Game Agents / NPC