Agent Environments

Revision as of 12:36, 22 January 2018 by Feld00 (talk | contribs)
Jump to: navigation, search

The Paper topic: Agent Environments

Class: 4IT496 System Simulation (WS 2017/2018)

Author: Bc. David Feldstein


In multiagent systems there are agents as programmed operating units in certain types of enviroments. The basic distribution of agents is reactive and delibarative agents. Reactive agents exist in the enviroment, they are influenced by its properties and changes however they do not create symbolic representation of the enviroment. Simply they don't try to simulate inteligent decisions or brain work. They don't read the environment, make no logic assumptions, they just react to it. In this case the inteligent behaviour comes from emergencies in the system which will be looked upon closely in below. On the other hand deliberative agents do try to simulate inteligence as we percieve it in our brains. They gather information from the environment and they do create symbolic representation of it. And based on their experience they try to make an adequate inteligent decisions.

This textbook chapter shows differences between different types of agents environments. It offers possible perceptions of the environment and how it affects each individual agent interacting with it and each other.


There are serveral different points of view in the environments which tells us how the environment can be percieved and helps us adapt the agents to operate in it.

Accessible vs. Inaccessible

– Is it possible to gather full and complete information about the environment? – Temperature inside this room vs. Temperature inside a volcano – The environment must be accessible to the agent in the moment. It is not enough that it is theoretically possible to get there.

Deterministic vs. Non-deterministic

– Does any action have a definite effect? Can we be sure about the state of the environment after the action? – Calculator vs. Room with a thermostat – Seldom it is so clear. Suppose a vending machine. It seems to be deterministic but what if a cup jams inside or the power drops out during the vending? Environment 9

Static vs. Dynamic

– Is the agent the only entity that changes the environment in the moment? Is it changing during the action of the agent? – Chess vs. Traffic Environment – Many real systems are dynamic, because many entities commonly act concurrently. In fact, there is no clear distinction between these two options. Suppose multitasking on a mono core computer.

Discrete vs. Continuous

– Is a number of possible actions in the environment finite or infinite? – Roulette vs. Legal system – Discreetness and continuity are often a matter of time. If a system is turn-based, it is discrete and often static as well. However, it could be dynamic when more participants make decisions together. 10

Episodic vs. Non-episodic

– Agents operate in certain segments (episodes) that are independent of each other and agent’s state in one episode has no impact on its state in another one. – Discrete simulation in Simprocess vs. whole economy – Distinction between episodic and non-episodic often depends on the viewpoint of the external observer.

Dimensional vs. Dimensionless

– Do spatial characteristics of the environment matter? – Real world vs. internet – If space, dimensions, distances, etc. play a role for agent’s decisions in the environment, then the environment is dimensional. If no of these characteristics are relevant, the environment is dimensionless.

Interpretation of the the environment


structured data for SW environments

Input function - can be from simple to very complex task (temperature - machine vision)

transduction problem


Environments recognition problem

Often, we cannot simply denote the environment either static or dynamic, either accessible or inaccessible, etc. The environment could have a certain level of the particular trait.

Environment complexity problem

The more inaccessible, non-deterministic, dynamic and continuous the environment, the more complex and less recognizable it is.

The more complex the environment, the more difficult it is to design an agent that should work there.

Time dimension of the environment

The agent is often constrained by time. Is cannot explore and analyze the situation for years, but it has to deliver results in a reasonable time.

short-term X long-term problem (especialy in dynamic environmet)

Subsumption architecture

Reactive agents architecture developed by Brooks. Two key ideas: – Situatedness and embodiment. The agents are physically present

in the environment, draw all their information from the interaction with it and directly influence environment’s dynamics.

– Intelligence and emergence. The intelligence does not exist per se. It emerges from agents’ interactions with the environment and it is not present in single specific component of the system

Agents sense the environment and their percepts directly trigger the proper actions. They are typically as simple as: situation → action.

• Situations are arranged into layers. The lower layer, the more specific behavior and the higher priority.

• The actions are fired concurrently, each layer has its own sensors and effectors.