A system which provides some simple worlds which are based on the style of the WOOD1 environment described from Wilson in "ZCS: A Zeroth Level Classifier System" (1994). Agents can connect with TCP sockets to these worlds and explore them. Objects are encode with the following ASCII symbols:
- . - Free space
-
-
- An Agent
-
- O - A Object which can't be passed
- F - Food
The socket connection use a plain text protocol, which is defined on the end of this Reamde file. The movement and position values are based on the following table where 0 is the actual position if the agent (also from Wilsons WOOD1):
8 | 1 | 2
7 | 0 | 3
6 | 5 | 4
Also compare with the corresponding script
http://www.uffmm.org/EoIESS-TH/gclt/node14.html
Disclaimer: The author of the script is different from the authors of
the implementations.
- Michael Bittorf (Coding Minds)
- André Freudenreich
- Min. Erlang R14B03
- Min. Scilab 5.4.0 with SOCKET library (2.0.1-1) OR
- Min. Python 2.7 OR
- Min. Common Lisp 2.49 with usocket (0.5.5)
- Replace PIDs with unique names or sthg like this. list_to_pid is only for debugging purposes and erlang interna.
- Build clients which correspond with the script
- Improve lcs_dynamic.py to build classifier list correct (diff. qualitys). Also ad handling of wildcards.
world (application)
'-> world_sup (supervisor)
|-> world_ofo_sup (supervisor)
| |-> world_logging (gen_server)
| |-> world_ctl_sservsup (supervisor)
| | '-> world_ctl_sserv (gen_server)
| |-> world_sservsup (supervisor)
| | '-> world_sserv (gen_server)
| '-> world_http
'-> world_envsup (supervisor)
'-> world_env (gen_server)
The global supervisor which initialize the whole application. First the one_for_one supervisor and then the simulated worlds.
A supervisor with one_for_one restart strategy which monitors all supervisor ans worker which can be restarted without big trouble or dependencies.
The supervisor which manages all virtual environments/worlds.
The gen_server which represents a virtual world and holds all the logic.
The socket server which handle incoming connections and forward them to
the simulated world instance of world_env.erl
Listens on the port which is defined in world.app as 'port' (default
4567)
The socket server which handle incoming control connections and forward
them to the simulated world instance of world_env.erl. Also creates and
destroys worlds
Listens on the port which is defined in world.app as 'ctl_port' (default
4568)
A simple http server based on the build in httpd. Servers a static page with a project description, a read only overview of the environments and some contact informations. The environment overview is based on js and refreshs every second.
A simple logging server which writes the messages from client and
environment changes/action into separate files.
The location of the files is defined in world.app as log_client, log_env
and log_info.
Some helper functions which mostly used in more than one module.
The system is initialized with three example environments which show all possible objects, except other agents.
To load your own environment connect to the control port (see Example usage) and use the commands 'world' and 'map' to load an existing environment for your connection and replace the existing map (This will drop all clients which are connected to this world). You can also spawn a new environment with a new world if you use 'world spawn [ASCII]'. Encode the map with the ASCII symbols described in the section Description; to mark the beginning of a new row use the character |. For example the defaultMap could be encoded with the following string:
.....|.....|OO.|...OF|..X..|
As you can see, the third row has only three elements. This should demonstrate, that the system interprets each not explicit defined section as a blocking object. And each unknown character (e.g. the X) is interpreted as a free cell.
Also the environment has some configuration options which are not direct tied to the map.
- maximum amount of allowed agents (default: as much as possible)
- should consumed food respwan (default: yes)
- should consumed food respawn on the same place (default: yes)
- the name of the environment (default: unknown)
- possibility to define the initial start position of the agents (default: yes)
- fitness if agent has not moved (default: 0)
- fitness if target section is blocked (default: 0)
- fitness if target section is staffed (default: 0)
- fitness if agent has moved (default: 0)
- initial energy (default: 1000)
- energy reduction if agent has not moved (default: 1)
- energy reduction if target section is blocked (default: 3)
- energy reduction if target section is staffed (default: 3)
- energy reduction if agent has moved (default: 2)
- drop agents if their fitness reaches 0 (default: false)
- time slice for interactions with the world (default: 0)
All this options can changed on runtime if you are connected with the control port. The following command snippet show the active options and then change the maximum client count to 7 and let the food respawn on random places.
options 104 max agents (max_agents): 4 104 respawn food (respawn_food): true 104 static food positions (static_food): true 104 environment name (env_name): defaultMap 104 allow start position (allow_startposition): true 104 fitness if agent not moved (fitness_nomove): 0 104 fitness if section blocked (fitness_blocked): 0 104 fitness if section staffed (fitness_staffed): 0 104 fitness if agent moved (fitness_moved): 0 104 initial energy (initial_energy): 1000 104 energy reduction if agent not moved (energy_nomove): 1 104 energy reduction if section blocked (energy_blocked): 3 104 energy reduction if section staffed (energy_staffed): 3 104 energy reduction if agent moved (energy_moved): 2 104 drop agents if their fitness reaches 0 (drop_agents): false 104 time slice for interactions with the world (time_slice): 0 104 EOL options 7 true false defaultMap true 0 0 0 0 1000 1 3 3 2 false 0 201 success
It's also possible to change only a single option
options 104 max agents (max_agents): 4 104 respawn food (respawn_food): true 104 static food positions (static_food): true 104 environment name (env_name): defaultMap 104 allow start position (allow_startposition): true 104 fitness if agent not moved (fitness_nomove): 0 104 fitness if section blocked (fitness_blocked): 0 104 fitness if section staffed (fitness_staffed): 0 104 fitness if agent moved (fitness_moved): 0 104 initial energy (initial_energy): 1000 104 energy reduction if agent not moved (energy_nomove): 1 104 energy reduction if section blocked (energy_blocked): 3 104 energy reduction if section staffed (energy_staffed): 3 104 energy reduction if agent moved (energy_moved): 2 104 drop agents if their fitness reaches 0 (drop_agents): false 104 time slice for interactions with the world (time_slice): 0 104 EOL options drop_agents true 201 success options 104 max agents (max_agents): 4 104 respawn food (respawn_food): true 104 static food positions (static_food): true 104 environment name (env_name): defaultMap 104 allow start position (allow_startposition): true 104 fitness if agent not moved (fitness_nomove): 0 104 fitness if section blocked (fitness_blocked): 0 104 fitness if section staffed (fitness_staffed): 0 104 fitness if agent moved (fitness_moved): 0 104 initial energy (initial_energy): 1000 104 energy reduction if agent not moved (energy_nomove): 1 104 energy reduction if section blocked (energy_blocked): 3 104 energy reduction if section staffed (energy_staffed): 3 104 energy reduction if agent moved (energy_moved): 2 104 drop agents if their fitness reaches 0 (drop_agents): true 104 time slice for interactions with the world (time_slice): 0 104 EOL
Notice: There are more options planed. E.g. the amount of food if an agent enters a food section, etc..
All ports and paths can be configured with ebin/world.app.
Before the first run (or after sourcecode modifications) you need to compile the .erl files. This can be done by hand
$ erlc --makeor with the include makefile
$ makeTo start the application start an erlang shell in the root of the project directory
erl -pa ebin/and enter
1> application:start(world).Then open a connection from another terminal and try to find the food
$ telnet localhost 4567 Trying 127.0.0.1... Connected to localhost. Escape character is '^]'. 200 Speak, friend, and ente(r) world list 105 ID, Name, X, Y, Agents, possible Agents 105 <0.97.0> restrictiveDemo 9 9 0 2 105 <0.99.0> WOOD1 55 15 0 1 105 <0.95.0> defaultMap 5 5 0 4 105 EOL world load <0.95.0> 200 welcome in this 5x5 world. Your ID is <0.56.0> environ 102 environ 0:OO...OOO move 5 201 success 0:...OOOOO move 3 201 success 0:....OO.. move 3 201 success 0:...O.O.. move 5 201 success 0:..OO..O. move 5 201 success 0:.OO....O move 4 201 success 0:OF.OOO.. move 2 202 food 1000:OOOO..OO quit 200 good bye Connection closed by foreign host.
If the server is running open a telnet connection to the control port
$ telnet localhost 4568 Trying 127.0.0.1... Connected to localhost. Escape character is '^]'. 200 Speak, friend, and ente(r) world list 105 ID, Name, X, Y, Agents, possible Agents 105 <0.97.0> restrictiveDemo 9 9 0 2 105 <0.99.0> WOOD1 55 15 0 1 105 <0.95.0> defaultMap 5 5 2 4 105 EOL world load <0.95.0> 107 World <0.95.0> loaded map 100 *1 *2 . . . 100 . . . . . 100 O O . O O 100 . . . O F 100 . . . . . 100 EOL map F...O|....|..........F 201 success 100 F . . . O O O O O O O 100 . . . . O O O O O O O 100 . . . . . . . . . . F 100 EOL quit 200 good bye Connection closed by foreign host.
A short Scilab socket connection demo.
A short Scilab program which connects to the server and let user interact with the world.
A simple demo of a random acting agent which terminates if he found
food. The agent loads the first possible world.
The amount of attempts will be printed to stdout.
A simple demo of a random acting agent which terminates if he found
food. The agent loads the configured or first possible world and
interprets food and objects to improve his behaviour.
The amount of attempts will be printed to stdout.
A java GUI client. The agent can be controlled manual or with a simple random based algorithm.
A short dynamic learning classifier system where the function which
applies the fitness values and the stored memory to the classifier list
can be simple switched. Also some options like world, startposition and
the size of the memory can be configured.
The amount of attempts per connection will be printed to stdout.
A simple demo of a php UI to show interested users what's going on. Uses the ctrl port to get the whole environment. (deprecated)
100 ASCII map representation
100 EOL
101 world changed (deprecated)
102 environ [0-9]+:[.|O|F|*i]{8,}
103 free text help replies
104 option listings
104 EOL
105 ID, Name, X, Y, Agents, possible Agents
105 ID X Y Agents MaxAgents
105 EOL
106 world ID spawned
107 world ID loaded
107 world ID destroyed
200 Speak, friend, and ente(r)
200 welcome in this [1-9]+x[1-9]+ world. Your ID is PID
200 good bye
201 success [0-9]+:[.|O|F|*i]{8,}
202 food [0-9]+:[.|O|F|*i]{8,}
203 blocked [0-9]+:[.|O|F|*i]{8,}
204 staffed [0-9]+:[.|O|F|*i]{8,}
205 failed
300 bad argument
301 dead
400 unknown command
403 access denied
404 not found
405 invalid position
500 sever made a boo boo
501 world destroyed
world list
world spawn ASCII_REPRESENTATION?
world load ID
world destroy ID
map ASCII_REPRESENTATION?
options (OPTIONS ..)?
drop dead
drop all
shutdown (not yet implemented)
quit
world list
world load ID (X Y)?
move [0-8]
environ
help [COMMAND]
quit
- Erlang documentation
- http://learnyousomeerlang.com/ for supervisor based socket server
- http://www.uffmm.org/ for theoretical stuff and specifications
- Twitter Bootstrap
- Source code: GNU General Public License v3
- Documentation: Creative Commons Attribution-ShareAlike 3.0 Unported License