Java程序辅导

ABS Assignment Lab 1
Rasmus Dall - r.dall@sms.ed.ac.uk
January 24th 2013
Introduction
In this lab you will be introduced to AgentSpeak and the Jason programming
language through the InfoSpeak environment. You will explore the environment
and learn how to define agents in Jason. The environment you will work with
looks similar to the George Square campus area - however it is not a faithful
representation. Please note that the environment is an experimental prototype
and as such you may come across imperfections and bugs - please report any
serious issues to me, I will attempt to fix them quickly if they are gamebreak-
ing. Before the lab you should all have read the provided chapter in the book
“Programming Multi-Agent Systems in AgentSpeak using Jason” - in the lab
this will be assumed and so I will not spend much time on the basics of Jason
but rather more time on familiarising you with the environment.
Setting up the environment
Jason is an interpreter for AgentSpeak programs, and a framework to create
environments for the development and testing of multi-agent systems. You will
not need to be familiar with the full power of Jason. In your assignments the
environment is provided and you need only create simple plan-rules that govern
the agent’s behaviour. The website for Jason is http://jason.sourceforge.net
and it is the best place to start for finding documentation and example code
(also see the link provided below). In particular the documentation section of
the site contains links to the manual and a “getting started” tutorial, both of
which are useful reference documents.
An IDE is provided for the easy editing of AgentSpeak programs, running the
environment and debugging. To set it up simply follow the instructions below,
taken from http://jason.sourceforge.net/mini-tutorial/getting-started/ (which
has more info, and pictures).
1. Download Jason from http://jason.sourceforge.net
2. Extract the archive to a folder in your userspace. Something like tar xzf
Jason-1.3.6a.tgz should do the trick.
1
3. Execute Jason by running ./jason.sh from the bin directory.
4. You will probably need to setup the location of your Java JDK. Go to
the menu Plugins → Plugins Options→Jason. On a DICE machine, ”Java
Home” should be set to “/usr/”.
Running a project
Let us test if the environment works by first running an example program.
• Open → jason-dir/examples/cleaning-robots/mars.mas2j
All Jason projects will be started using the .mas2j file, this files defines the
environment and agents in it. In this case the MarsEnv and the agents r1 and
r2. Each of these will have seperate definition files (MarsEnv.java and r1.asl,
r2.asl).
• Run the program by pressing the “Run MAS” button in the lower right
hand side.
A window will pop-up and a simulation of the mars robots you should have
encountered in your lectures will run. If it runs your system should be correctly
set up. If not let me know so we can fix the problem.
The InfoSpeak world
The InfoSpeak environment which you will be working with is an example of
a grid world agents can operate in. The grid depicts a (putative) map of the
university campus where white squares are in-accessible, and agents (coloured
ovals) move around the shaded squares. Squares labelled OUT are used to
denote leaving the campus, LT buildings are lecture theatre locations, FOR
denotes the Informatics Forum, LIB the University Library, LAB a laboratory
location, BAR a pub students like to go to, and GYM a gym. The green block
of squares in the middle depicts a park. Agents are capable of moving across
squares, and they can perceive their locations and find routes to locations. The
“Select Agent” menu will display an agent’s position. A right mouse click on a
building will block it for access by agents (red color means “blocked”), this can
be undone by right-clicking on it again. To set up the environment:
1. Download the code from the course website and decompress it, e.g. with
tar xzf abs1.tgz.
2. Start Jason, and load up infoSpeak.mas2j.
There is also a variety of other files for the assignment - of these you will
be expected to use the infospeak.config file and to define the two agents
focusedAgent.asl and chaoticAgent.asl. The provided agent lecturer.asl
may be useful to study.
• Try running the environment as setup with one lecturer agent.
2
The .config file
The simulation environment expects a configuration file infospeak.config
which is used to read in agent types, agents, agent groups (which can be defined
to share certain elements), buildings, assignments, lecture details, and custom
events on startup. Note that the number of agents specified in infospeak.mas2j
needs to be identical to those agents specified in the configuration file. The syn-
tax of the configuration file is self-explanatory and you will be expected to
utilise this in your assignment. Not all the elements of the system that can be
configured are necessarily used in the assignments.
A Brief Overview of an AgentSpeak Agent
This is a very brief overview, for further information you are expected to read
the provided chapter in the book. The main components of an agent are beliefs,
intentions and plans. Beliefs either come from percepts, and are updated auto-
matically each reasoning cycle, or are added from plans. They are predicates,
which may be of arity 0. An example of a position belief might be pos(2, 3).
Plans are of the following form:
triggering-event : context <- plan_step; plan_step.
Triggering-event - Defines which events may initiate a plan. It could be
internal, from plans, or external - from percepts. It can be addition,
denoted by + or removal denoted by -. It may be either a goal or a
belief. So for example +!start would trigger on the addition of a start
achievement goal and +winter(isComing) would trigger on the addition of
a winter(isComing) belief.
Context - The context in which a plan applies. A conjunction or disjunction
of beliefs, logical formulae and certain built-in special predicates, such as
.random(N). An example of a context depending on the position would be
pos(X,Y) & (X==10). Note that variables must begin with an uppercase
letter, lowercase or numbers are constants. ”_” unifies with anything, and
is the ”don’t care” symbol.
Plan_step - There may be 0 or more steps, separated by ; and terminated
with a full-stop. These can be actions, of the form do(Action), goals to
achieve (of the form !goal(_)) or the addition or deletion of beliefs (of
the form +winter(isComing) or -winter(isComing)).
Goals take the form !goal(_) and like beliefs may have an arity of 0.
Percepts
Percepts provided by the world are as follows and are automatically added to
the agents belief base:
3
pos(Object,X,Y) - The object Object (a building or an agent) is located at
square (X, Y).
home(X,Y) - The agent’s home location (one of the OUT squares)
day(D) - The current day (an integer counting days since the start of the
simulation)
time(T) - The time of the day (a number between 9 (a.m.) and 24 (12 a.m.)
denoting the hour of the day, “nighttime” is simulated as a single time
step so the time jumps from midnight to 9 a.m.)
week(W) - The current week of the term (an integer)
lecture(Coursename,Week,Day,Time,Place,(Priority)) - Denotes that course
Coursename takes place every Weekday at time Time with a certain Priority
(this specifies which lecture would be preferred in case of clashes and is
optional)
assignment(Course,N,SWeek,SDay,STime,EWeek,EDay,ETime,Hours)
- Denotes that an assignment that requires Hours hours of work runs from
STime on SDay in week SWeek until ETime on EDay in week EWeek
event(T,SWeek,SDay,STime,EWeek,EDay,ETime,Location,(Priority))
- Custom event of type T (e.g. party, job, lecture, interview) with start and
end times/days/weeks, occurring at Location with a certain Priority
(the last argument is optional)
Actions
Besides the standard actions (which are of the form .action(_) - remember the
dot) this particular environment has the following additional actions:
go_to(X,Y) - Move toward square (X,Y)
one_hour(Time) - Spend one hour while remaining in the current location
add_goal(Text) - Set the goal of the agent to be displayed to Text in the
GUI (does not affect actual goals pursued, only for visual display)
For the second assignment a number of additional actions are available:
revise(Time) - Revise for one hour in the current location
attend_lecture(Time) - Attend a lecture
attend_event(Time) - Attend a custom event defined in the config file for 1
hour
attend_meeting(Time) - Attend a meeting between agents for 1 hour
4
work_on_ass(Time) - Work for 1 hour on an assignment
submit(Time) - Submit an assignment (takes 1 hour)
Note that the additional actions for the 2nd assignment are not particularly “in-
telligent” they do not e.g. check if the agent has actually finished an assignment
when submitting. These are for scoring purposes only and may be used inap-
propriately in the 2nd assignment - make sure you only call these at appropriate
times in your solution. I.e. you will not receive marks for agents which use the
submit action when it has not finished an assignment.
Unification
One important difference between Jason and other Prolog-like languages (i.e. in
logic programming) from conventional programming languages such as Java or
C is the concept of unification. In Jason when we write e.g. pos(Object,X,Y)
then Object, X and Y are all variables which needs to be instantiated or unified.
Where in e.g. Java when calling a method like pos(Object,X,Y) we must provide
the method with the values for Object, X and Y - we can in fact in Jason use
this to retrieve the various variables. So in the infospeak environment when we
write time(Time) then ’Time’ will unify with whatever value the belief time(_)
has in the agents belief base such that if the agent has the belief time(9) then the
value of ’Time’ will be 9. Should the agent have no such belief the variable will
remain uninstantiated and have no value. Consider this plan from the lecturer:
+!givelectures(Me) : week(WeekNow) & day(Day) & time(TimeNow) & pos(Place,
X, Y) & lecture(Name, Week, Day, Time, Place, Priority) & Time
== TimeNow & (Week == 0|Week == WeekNow) <- add_goal(“Going to
the next lecture”); !goto(Me,X,Y); .concat(“Attending “, Name,
” lecture”, G); add_goal(G); one_hour(Time); !checkevents(Me).
The only variable which is unified from the beginning is Me which is unified
with whatever value Me in !givelectures(Me) takes when it is added as a goal.
For the rest these are unified along the way, week(_), day(_) and time(_) each
unifies with the value of the current week/day/time which can then be used
later. pos(Place,X,Y) unifies Place, X and Y to the current position of the
agent. Let us say that in this case Day == Monday, once we get to the lec-
ture(Name,Week,Day,Time,Place,Priority) this will retrieve any beliefs in which
the Day of the lecture is Monday and will then unify the rest of the variables
according to the lecture found; so if the student had a lecture “abs” on Monday
in week 7 at 11 o’clock in at1 with a priority of 2 the total resolution of lecture
would be lecture(abs, 7, monday, 11, at1, 2) and these values can now be used
further ahead. In this case it is checked if the current time matches the time
of the lecture and the same for the week, if such a lecture is found the plan
is then executed. Remember that variables all start with a capital letter and
atoms (the actual values) are all lowercase. Atoms can also be used in contexts
e.g. in
5
+!gotoforum(Me) : pos(frum1, X, Y) & pos(Me, Xm, Ym) & X == Xm & Y
== Ym.
The atom ’frum1’ is used which ensures that X and Y will unify with the x and
y values of the frum1 building.
Playing around with the environment
Try playing around with the lecturer agent and the config file try e.g.:
• Adding several lecturer agents and assign them differing courses, events
etc.
• Blocking a building (right-click the building) that an agent wants to enter
• Changing the number and names of courses and/or events
• Running an agent for a few hours - stopping it and inspect its debug
window
• Changing the name and location of a few buildings
• Making agents go to non-existent buildings
• Changing the speed of hours passing
• Changing the lecturers priorities of events and courses and make a clash
happen
• Making the lecturer go home from the forum later/earlier/work in the
library instead
• Making the lecturer excercise in the gym instead of going to lunch
• If you have time you can start to work on the assignment
A infospeak.config.BACKUP file is provided which you can use to reset the
environment should you need it. I will be available to answer questions and the
like for the rest of the lab while you try out various things.
Idiosyncrasies to be aware of and a few hints and
tips
• When working on the assignment you may find the following link useful -
http://agentgame.mit.bme.hu/files/Jason/Jason_konyv.pdf
• The InfoSpeak environment does not deal well with large numbers of
agents (above 5 or 6). There is no need to run more than 4 (or less)
in your assignments.
6
• If you have paused the environment in order to debug do not resume the
simulation. The environment does not deal well with this. Close the
simulation and restart it instead.
• The number of agents and names of agents defined in the .mas2j file and
.config file must match.
• If the context of several plans applies the first plan the interpreter encoun-
ters will be executed (ordering is important)
• An agent which is executing a plan with several plan_steps will not drop
later steps automatically if an earlier step adds additional steps to do -
rather it will do the newly added steps first and then resume executing
the later steps. If you wish you can reset all agents intentions, desires
etc. using the actions .drop_all_intentions or .drop_all_desires
(the above provided link may help in clarifying this).
• While it may be useful to take a look at the ISEnv.java file it should
not be necessary for you to change it, if you feel for some reason this has
become necessary - you’re either doing something wrong or there’s a bug
I should know about.
7