In this assignment, you will implement some simple soccer-playing agents. All parts of this assignment can be done with purely reactive behaviors. Please describe in your email to us the extent to which your behaviors are reactive. (It's OK if they're not, just make sure that if you create an agent that is non-reactive in some way, you describe clearly why not in your email).

This assignment (and all for this course) is about creating agents. It's not about programming. If you have programming questions/problems, etc. (e.g. "How do I generate a random number in C?") ask the class list early and often.

A good debugging tool is to print stuff out, in C you can do it with printf or in C++ you can do it with cout. Here is a simple example which should help (for more info check the printf reference):

  #include <stdio.h> // This should be at the top of the file
  ...
  char str[10] = "agents"; 
  int num = 1;
  double dbl = 2.0;
  // It prints out the value of variables str,num and dbl. 
  // Notice the difference between %s, %d and %f.
  printf("string: %s, int:%d, double:%f\n", str, num, dbl); 

The codebase also has a LOG_STR macro which takes in a string and prints out both the string as well as the line of code the string is being printed from. The code LOG_STR("hello"); placed in selectSkill() at line 82 of strategy.cc would produce the following output: strategy.cc:selectSkill():82:: hello.

In RoboViz you can draw things on the field to help with debugging. The following is an example of how to draw a red dot on the field 1 meter away from the agent and 15 degrees from the agent's current orientation.

  char buf[512];
  sprintf(buf, "DrawPoint.%d", worldModel->getUNum());
  string str = string(buf);

  float pointDistance = 1.0;
  float pointAngle = 15.0;

  // Convert from polar to cartesian
  VecPosition drawPoint = VecPosition::getVecPositionFromPolar(pointDistance, pointAngle, 0); 

  // Converts point from local to global coordinate system that
  // RoboViz uses when drawing.  This conversion is possible as the
  // agent localizes itself using a particle filter (or in the case of
  // this assignment ground truth localization data provided by the server).  
  drawPoint = worldModel->l2g(drawPoint);

  // Need to rotate point 180 degrees around origin to line up with global
  // coordinate system if agent is starting on the right side
  if (worldModel->getSide() == SIDE_RIGHT) {
    drawPoint = drawPoint.rotateAboutZ(180);
  }
  
  // Set color of dot
  float red = 1.0;
  float green = 0.0;
  float blue = 0.0;

  // Set size of dot
  float size = 5.0;

  // Set position of dot
  float x = drawPoint.getX();
  float y = drawPoint.getY();
  float z = 0.0;

  // Write drawing code to str
  worldModel->getRVSender()->drawPoint
    (
     x, y, z,
     size,
     red, green, blue,
     &str
   );

  // Pass str containing draw instruction to RoboViz
  worldModel->getRVSender()->swapBuffers(&str);

A good method for parsing strings, is sscanf. Here is an example:

  #include <stdio.h> // This should be at the top of the file
  ...
  char str[20] = "((b) 23 1.5)"; 
  int i = 0;
  double dbl = 0;
  // The integer value appeared at the place of the %d would go to the i variable, 
  // and the the double value would go to the dbl variable.
  sscanf(str, "((b) %d %lf)", &i, &dbl); 
  printf("i:%d dbl:%f\n", i, dbl); 

The strstr() function may also be useful. (Here is code that parses a sample server message and find the distance and angle to the ball.)

Also, the sprintf() function is useful for creating strings with variable content. For example, you can use sprintf() to create a

(dash N)

C/C++ Reference

Score a goal

Instructions:

1. Copy over the files in the nao-agent directory to any directory in your account:

     % cp -r /projects/cs344M.pstone/3d/nao-agent .
   

   IMPORTANT: You will need to do this step again even if you already did this for the first assignment (you should delete all files copied from the first assignment) as the codebase has been updated for the second assignment.
2. Go into your version of the nao-agent and compile by typing "make".
3. In a new terminal window, start the soccer server (as you did in the previous assignment).

     % rcroboviz
   

4. Back in the other window, run the agent:

     % ./agentspark --host=localhost --team myteam --unum 2 --paramsfile paramfiles/defaultParams.txt&
   

You will see the same behavior as last week. This time instead of starting play with a kickoff do a drop ball by pressing 'b' or press 'o' and choose PlayOn from the play mode selection menu. This is necessary as your agent will want to touch the ball multiple times during this assignment and kickoff rules limit the agent to only touching the ball once until another agent touches the ball after a kickoff.

Your task is to alter the code in strategy.cc that the player does the following :

• Immediately after connecting, it should move to a random place on its side of the field (set values in the beam() method) where -15 < x < 0, -10 < y < 10.
  
  
• It should find the ball "(B", go the the ball, find the opponent's goal posts, "(G1R" and "(G2R" for the left team or "(G1L" and "(G2L" for the right team, and kick the ball towards that goal. The player should decide what to do based only on its latest sight.

• To determine which side a player is on use the worldModel->getSide() method which returns an enumerated value of either SIDE_LEFT or SIDE_RIGHT.
  
  
• Parse the string returned by the worldModel->getLastVisionInput() method to find the distance and angle to the ball and/or the relevant goal posts (if they're in the player's view).
  
  
• The format of visual information for seeing an object is:

  (ObjName (pol <radius> <theta> <phi>))
  

  These values are in a spherical coordinate system but you don't really need to understand this as the VecPosition class makes it easy to convert from spherical to cartesian coordinate values. Also it is a good idea to transform these measurements from the robot's camera to be relative to its body origin for better accuracy. Look at the code again in tst.cc to see an example of how to do this. The ObjName variable can be many things, including:
  B for the ball
  G1R and G2R for the top (left) and bottom (right) right side of the field goal posts respectively
  G1L and G2L for top (right) and bottom (left) left side of the field goal posts repectively
  Look at the Field Dimensions and Layout in the online Simspark Users Manual for a diagram of where the goal posts are.

  Note that objects can only be seen in a 120 degree view cone.

• To walk return the value from calling the getWalk(direction /*in degrees*/, rotation /*in degrees*/, speed /*0-1*/) method.

     getWalk(0, 0, 1); // Walk forwards
     getWalk(180, 0, 1); // Walk backwards
     getWalk(-90, 0, 1); // Walk to the right
     getWalk(90, 0, 1); // Walk to the left
     getWalk(0, -90, 0); // Turn clockwise
     getWalk(0, 90, 0); // Turn counter-clockwise
  

• To determine if the agent is close enough to the ball to kick it call the canKick() method and see if it returns true.
  
  
• When the agent is close enough to kick the ball you can then call and return the value from the kickBall(kickTypeToUse, direction, distance) method. This will put the agent in an auto pilot mode to kick the ball in the specified direction. Parameters for kickBall are the following:

     kickTypeToUse: enumerated int value of KICK_LONG, KICK_MED, KICK_QUICK, KICK_IK, or KICK_DRIBBLE (any of these are fine)
     direction: angle in degrees relative to agent's current orientation to kick the ball toward
     distance: distance away from the agent that its desired target to kick the ball toward is
  

• If you want to cancel kicking the ball for whatever reason you can call the stopKickBall() method.
  
  
• Sometimes the robot will thrash or "dance" around for a bit while trying to position to kick the ball. Also the robot will lose its balance and fall over from time to time as well as miskick the ball. These behaviors are OK and something we're working on to try and improve (later on in the class you can look into improving these behaviors if you like).
  
  
• A good goal scoring agent can sometimes score 5 or more goals in a half with a little luck.

What to turn in:

1 on 1

What to turn in:

A logfile called [yourlogin]-1on13d.log compressed with gzip.

Passing

Hints:

(P (team <team_name>) (id <jersey_num>) (<BodyPart> (pol <radius> <theta> <phi>)) (<BodyPart> (pol <radius> <theta> <phi>)) ...)

Grading

This assignment will be graded as follows:

Part 1:

• 4 points if it works smoothly and scores twice in a half.
• 3 points if it doesn't work so well but still scores

Part 2: 1 point if it works.

Part 3:

• 4 points if it works smoothly. (it's OK if the ball doesn't go straight to the teammate, as long as the teammate can go retrieve the ball and keep playing)
• 3.5 points if a couple of passes are completed, but it doesn't seem like it could keep working (players get bunched up, etc.)
• 3 points if 1 pass is completed.

Description:

• 1 point if you indicate an understanding of "reactive"
• .5 if you provide a poor description
• 0 if you don't provide a description