next up previous contents
Next: EMuds, Aims and Methods Up: EMuds Previous: Artificial Intelligence and Agent


Games and Text-based Virtual Reality

When the AXE project was formulated, the main idea was that the problems of global networking can be solved by using populations of agents distributed in the network. This vision is based on the fact that agents living in a virtual world made of interconnected computers should naturally be able to solve questions of distributed problem solving, network maintenance or load balancing by adapting them to perceive the properties of their current location and learning how to move in the network, possibly picking up data from one location and transferring it to another location. With this goal in mind, an experimental platform that could support these ideas was deemed necessary and it is this platform that was designed in the EMud environment model. In order to think about distributed networks of computers, a game metaphor was the first idea that occurred to me, inspired by my role-playing background. It is this gaming environment concept that I introduce in this chapter by showing the evolution of computer games from their beginning to the current Mud games.

The Origins of Computer Games

The evolution of the use of computers for gaming purposes has basically gone through three stages. At first, computers were used as a replacement for a human partner or opponent in a classical game that was until then played without computers. This is typical of the strategy genre of games, which are usually of the board game type, but transposed to computers.

In the next stage, games were developed specifically for computers and these cannot easily be played without a computer base. Two genres represent this evolution: action games, where the player must match his dexterity against a series of challenges generated by the computer, and the adventure kind of games where a player wanders around in a virtual setting described by the computer (textually or visually) and issues short command sentences to try to solve enigmas. Early representatives of these game families are typically chess or Othello (Reversi) for strategy games, Pacman or space invaders for action games and Advent or Zork (Dungeon) for adventure games.

Although these categories can be further divided and are slightly mingled in more recent games such as wargames, simulators or computer role playing games (CRPG) of which we will speak later, the last transition in computer game concept was introduced in the late seventies with the use of networking for online gaming. Technically, online games are not a new genre, but are rather an extension from all other genres to human-human interaction through the computer media.

Although online games have been available since the seventies, it is only during the last few years that their success has been rising sharply. Until recently, there were no commercial games of that type, which can be explained by the limited popular impact of such a product in a world without widespread Internet access. But today, it seems that the ``multiplayer'' feature will be inescapable for any game in the near future. Indeed, a large part of the success of games such as the shooting game called Quake is that they can be played cooperatively or competitively with other players over the Internet. Maybe one could consider this as one of the failures of artificial intelligence techniques used to generate ally/enemy behaviors for games where the rules are more numerous than in classical board games, but in any case, the challenge of comparing one's skill to that of other human player will always remain.


From Fantasy Literature to Advent and Computer Role Playing Games

The publication of ``The Lord of the Rings'' in 1937 [62] marks the beginning of fantasy literature by using typical mythological and epic elements in a modern fiction novel. A large community of fans gathered around this book and its intricate description of the fantasy world in which the story takes place. Much literature has since then been published with a similar setting, but the importance of the genre is also due to its propensity to produce the attractive imaginary worlds that attract players of role playing and adventure games.

In 1974, Gary Gygax releases his game system called Dungeons & Dragons (D&D), Basic Set [21], which is the first role-playing game (RPG) to be published. In D&D and role-playing games in general, players design a character based on the game system's rules and writes this description on a character sheet. A referee (the game master or GM) invents a story with descriptions of puzzles, opponents, treasures all fitted in a world setting that he has imagined. The players and the GM then meet around a table to ``live'' through the scenario created by the game master, who acts as an omniscient and omnipotent entity ruling out the results of the behaviors of the characters in his world. Role-playing games now come in a wide variety of settings, from fantasy and modern times to science fiction worlds, but D&D was distinctly based on Tolkien's world of elves, dragons and other mythological creatures.

Actually, this type of games is quite similar to role-playing experiments in psychology [41] except that players do not act out their chosen behavior, but explain it to the referee who then describes the results. (Which is probably a good thing since role playing games often involve quite a lot of hacking and slashing at monsters with swords!)

It is in this context that in 1976, Will Crowthers designed Advent (the six character name restriction, that was then common place on computers, for adventure), a computer refereed fantasy game inspired both by fantasy literature and his own (real) caving experience. The game was subsequently greatly expanded and released by Don Woods and was the first adventure game [25]. In Advent, the player explores a text-based description of a cave system, populated with creatures and objects, where he must solve puzzles to reach deeper into the cave system and score more points. Shortly after the appearance of Advent, another game called Zork [34] (later called Dungeon for its Fortran version) was written along the same line by Tim Anderson, Marc Blank, Bruce Daniels and Dave Lebling in a language called MDL.

Playing Adventure Games

In a text-based adventure game like Advent, the player acts as if he was a person in an imaginary world, this person is usually called his character in the game. He is given textual descriptions of what that person can see around the place he is standing in the adventure world such as:
You are standing at the end of a road before a small brick building. Around you is a forest. A small stream flows out of the building and down a gully.
Which is the first screen of the Advent game. The player may then issue various commands that his character will attempt to execute. A command usually consists in an action verb followed by the object of that action. After mucking about a little, the player might try to enter the described building:
> enter building

You are inside a building, a well house for a large spring.

There are some keys on the ground here.
There is a shiny brass lamp nearby.
There is food here.
There is a bottle of water here.

And pick up the objects he sees:
> get everything

Sorry, I don't know the word ``EVERYTHING''.

> get lamp


> get keys


When commands are not understood, the game will inform the player and ignore the attempts. Otherwise, various humorous or gruesome things can happen to a character playing Dungeon, the player might encounter the following situation:
You are in a dark and damp cellar with a narrow passageway leading east, and a crawlway to the south. To the west is the bottom of a steep metal ramp which is unclimbable.
The door crashes shut, and you hear someone barring it.
Your sword is glowing with a faint blue glow.

> inventory

You are carrying:
A sword.
A lamp.
A newspaper.
A leaflet.

> east

You are in a small room with passages off in all directions. Bloodstains and deep scratches (perhaps made by an axe) mar the walls. A nasty-looking troll, brandishing a bloody axe, blocks all passages out of the room. Your sword has begun to glow very brightly.

> hit troll with sword

The fatal blow strikes the troll square in the heart: he dies. Almost as soon as the troll breathes his last, a cloud of sinister black smoke envelops him, and when the fog lifts, the carcass has disappeared.
Your sword is no longer glowing.

Which is the happy outcome, the other would signal an end game...
You are in a small room with passages off in all directions. Bloodstains and deep scratches (perhaps made by an axe) mar the walls. A nasty-looking troll, brandishing a bloody axe, blocks all passages out of the room.
The flat of the troll's axe hits you delicately on the head, knocking you out.
Conquering his fears, the troll puts you to death.
It appears that the last blow was too much for you. I'm afraid that you are dead.
Do you wish me to try to patch you?

> no

What? You don't trust me? Why, only last week I patched a running RSX system and it survived for over thirty seconds. Oh, well. Your score is 35 [total of 585 points], in 14 moves.
This gives you the rank of Amateur Adventurer.

The aim of these games is to collect the objects that are lying around the various places in the imaginary world, finding those that serve to solve riddles allowing further progression and bringing home the treasure objects that increase the character's final score. Part of the fun is in discovering how to solve the riddles or avoiding the evil monsters that lurk in the darkness and part comes from exploring the world that the designer has written. In the examples that I have given it is not immediately apparent, but the syntax of these games is very restricted and a good deal of patience is also necessary to find the right formulation for an action that the character must execute but ``doesn't understand''.

The MUD and Muds

Inspired by Advent, Zork/Dungeon and in a lesser way by Hack (another adventure game), Roy Trubshaw at Essex University decided to write a multi-player adventure game in spring 1979. After writing a first basic assembler version where people can move around rooms and chat together, he rewrites it completely in a higher level language (BCPL). With Richard Bartle now doing much of the game play oriented programming, they finished in 1980 what is now believed to be the first MUD [4]. The name MUD is the acronym for Multi User Dungeon, where Dungeon comes from the Fortran name of Zork that inspired the authors of MUD. Nowadays, MUD is also sometimes taken to mean Multi User Dimensions since it is a more accurate reflection of the variety of settings (dimensions) in which one can play Muds.

The two ideas that brought R. Trubshaw to write the MUD were his interest in writing a database definition language and the making of a multi-player adventure game. It is the latter aspect that became most important over time and when he left Essex University, R. Bartle took over the project and added most of its game oriented features, that is, puzzles to solve, atmosphere, etc.

Popularized by students throughout the UK, MUD quickly spreads to Norway, Sweden, Australia and the USA. Since then, there have been many programs written either over the original MUD code or strongly inspired by it. These kernels form the basis of today's Muds, examples of these are typically LPmud or DikuMUD, and most current running Muds use these as a code base.

Playing Muds

The player of a Mud is faced with the same environment as that of an adventure game player. He receives textual descriptions of his immediate surroundings, including the objects the he may pick up or the enemies he can fight. There are also various simple riddles that he may solve in order to progress in the game and the commands that can be executed take the same form as in an adventure game. Where differences appear is in the characters that are played and the scoring system. A Mud character can choose among various professions and has a more detailed description. His skills can vary depending on the choices the player makes while playing and may have access to commands that not all other players are allowed to use, for example, thief-like characters might be allowed to steal objects from creatures and wizard-like characters may have access to magic spell casting. A multiplayer game character is in general very customizable, allowing players to decide how they look like, what their name is and what clothes they are wearing.

When exploring the world in a multiplayer game, the player's character encounters creatures in the same way as in an adventure game, but might also come across other characters. He can then engage in conversation by issuing the say command, followed by a sentence the he wants the other player to hear (see, in fact), or even make his character act in various visual ways by using the emote command to show the other player his emotions:

> east

You enter the central courtyard in Tintagel castle. Big double pane doors lead to the round table room in the north. To the south and west, secondary entrances lead inside the castle. In the middle of the courtyard stands large podium where King Arthur can address his people.

A barrel of water is here
Some hay is lying here
A horse is standing here
You see Vania

> look Vania

You remember having seen Vania at the court once, she must be a minor lord from the country. She is a mature woman of forty years who clearly has seen rough times. Her clothing is correct, but obviously designed for riding, and while her stance is proud, she could easily pass for a commoner.

She is carrying:
a leather backpack
a sword
a suit of leather armor
black riding boots

> say Hello Vania

You say ``Hello Vania''
Vania nods

> emote bow Vania

You bow before Vania

Depending on the Mud, mutual aggression may be possible or not, but it is generally considered bad taste and most Muds have a board of internal regulations to explain the behaviors that are allowed, tolerated or forbidden. These rules are enforced by special characters that are played by the maintainers of the Mud or simply people that have been playing for a long time and are deemed responsible. A lot of the reputation of various Muds that can be played on the Internet comes from the general atmosphere that is maintained for the players.

The scoring system in a Mud differs considerably from that of other games and is the same as in role playing games. In a Mud, characters are given an experience point score. When creating a new character, the score starts at zero and it is by fighting creatures (killing them actually) or solving quests that the character can gather experience. These points are used to calculate the level of advancement of the character in his profession and by accumulating them, the character can increase this level. With higher levels, character gain proficiencies in various areas dependent on their profession and the choices the player makes. All in all, more advanced characters become more powerful, allowing players to compare their characters according to what each can do or which monsters they are able to beat.


While early Muds emphasized competition, riddle solving, character development and can be called adventure Muds, so-called social Muds have since then been designed. These Muds emphasize creativity, programming skills and communication in the sense that their worlds are not created with populations of dangerous creatures, quests or generally hostile environments, but exist for the players to modify and enlarge by themselves. Characters on social Muds are allowed to build new places and new objects to interact with through the use of a programming language. Their aim is to design places of their own, that other characters can visit, admire and/or criticize. On such a Mud, the time not spent building is spent on communicating with other players or exploring their creations. The first decisively socially oriented Mud, called TinyMUD, was written by Jim Aspnes in 1989. It allowed players to create new object, places, etc. by spending pennies (Mud cash) acquired on the Mud. All characters had equal powers and the idea was that TinyMUD had to be a place where non-competitive people were happy to be. Following TinyMUD came TinyMUCK (the programming language used for creations involved less MUCKing about than on TinyMUD), TinyMUSH and many others who inherited the MUCK or MUSH nickname. Following the same basic concepts, MOOs were written to implement socially oriented Muds where building is done through the use of and object oriented programming language.

Figure: A historical hierarchy of Muds [29].

Text-Based Virtual Reality

Today, virtual reality is everywhere and the text-based virtual worlds of early games appear like the poor fathers of their fully animated, three dimensional, sound enhanced children that can be bought in every superstore. But, taking away the gore, every element that is truly entertaining was already in these first games: incredibly large worlds to explore, the challenge of tricky quests to solve, rewards for the (play) efforts involved and mainly places to live out one's imagination. These are even elements that now are often overlooked by the entertainment industry in favor of the thrill of new sensory experiences, even though their comeback can be felt is some recent games. This additional sensory experience that appears so important nowadays is a trend that appears in all computer-human interactions in the attempt to interface man and machine. But if we consider artificial agents (not interacting with humans, that is), this is probably the element that can most safely be omitted from the environments designed for them. Instead, an environment customized for an agent would more appropriately be implemented as a discrete symbolic world, since this agent will have a discrete symbolic internal mechanism.

In an attempt to ground symbolic agents in an environment, I believe that our first attempts should be concerned with environments that are intuitively close to the representation mechanisms these agents possess, making them feel at home in a sense. Once our methods enable such agents to solve complex problems in these environments, the next step will be to bring them to real world problems that anyhow must be converted to a discrete description for the agent perceptions. The idea that only real world problems can bring agents to integrate high level behavioral patterns, appears misguided to me. The real difficulty lies in an agent handling situations of a structural complexity level larger than its internal representational structure, whereby the agent must discover or be provided with adequate generalization/compression capacities in order to solve the problems he is faced with. I have chosen to use the EMud environments introduced in the next chapter as experiment environments for this reason. An EMud environment is a virtual text-based world designed from the principles of Muds. In this world, places and objects are of a symbolic nature, the transformations in this world are rule-based and so, the world can be directly experienced by an agent placed in it. Additionally, since an EMud is a Mud, human players can control an agent in the virtual world, allowing interaction with artificial agents. For these reasons, both complexity and the basic unpredictability of human actions can be introduced in the environment to provide a sufficiently challenging problem space for agents.

next up previous contents
Next: EMuds, Aims and Methods Up: EMuds Previous: Artificial Intelligence and Agent
Antony Robert