CS 474/574: Autonomous Systems

Documentation Page


Nserver manual

Nomadic's manual for the Nomad200 robot includes a section on Nserver. If your web browser is configured to handle Postscript files, you can view it right here. Otherwise, there's a copy in the /doc directory of the CS474 account; run the Postscript viewing program ghostview to see it.

Using Nserver

The manual in the /doc directory of the CS474 account document how the Nserver program supplied by Nomadic Inc. operates. For the purposes of this class, I've had to modify that operation somewhat. The command-line syntax of our server is:

 Usage:
 Nserver [-help -noisy -port ]  

   where options are:
      -help:   Print a short help message
      -noisy:  Load a "noisy" robot setup
      -port:   Force a port setting

    robot.setup and world.setup are optional files which are
    used as the robot and world setup files.
 
    Note all options can be abbreviated to the first letter.
Note that there are TWO different robot setups --- an "ideal" robot which has perfect sensors, and a "noisy" robot which has random noise associated with sensor readings. By default, the robot is "ideal."

Note that the robot and world setup files are optional.

Thus, if you just type Nserver (assuming you've got your path set up right), you should get a vanilla, idealized robot.

Running remotely

Since this is a client-server model, it is also possible to run Nserver on the Linux machines, but do your programming somewhere else (assuming you recompile the appropriate libraries). If you do, you have to do a couple of things:
  1. Create a way for the server and client to connect on the same port, presumably by copying the .Nserverport file (see the next section).
  2. Tell your client where the server is. This can be accomplished by setting the environment variable SERVER_MACHINE_NAME in your environment.

A couple of details

A little inspection will quickly tell you that what I've done is to use a shell script to run Nserver. Why? Well, Nserver uses a client-server model, with the connection between the two occurring over a fixed TCP/IP port (7019) by default. This is a problem since, it is not possible for more than one user to run Nserver. So, I've made up a scheme for dynamically allocation port numbers. The port flag bypasses this scheme and "forces" the use of a particular port. In order to make this happen, the following games are played:
  1. The script looks for a file, .Nserverport, in your home directory. If one is there, it assumes that you are already running Nserver and quites.
  2. Otherwise, it tries to find a free port. This is done by looking in /tmp on your machine to see if there is a lock file indicating someone is using a port.
  3. Assuming a port is found, the script runs Nserver with that port number, and also write a .Nserverport file.
  4. When you run your program, it looks for .Nserverport and uses the port number contained in it to establish the connection to the server.
Obviously, there are many ways for this scheme to go wrong if the appropriate files are note deleted at the appropriate times. Hence, understanding this scheme will help you fix up problems, should they occur.

The Sout Robots

See the scout howto page.
Copyright (c) 1997 Greg Hager, Dept. of Computer Science, Yale University 
 Last modified Jan. 26, 1997.