• Tom strongly urged we use Corba as a mechanism to communicate between AMC and VCS (i.e current Layerd Control software).  General agreement on its use, and its dependancy on a Linux port of the Layerd Control system was noted.
  • There will likely be an ORB which will be on the MVC and AMC/autonomy will connect to it.
  • Some complexity will increase with use of CORBA; however debugging is less of an issue since will have a similar issue with sockets.
  • Determined that sockets or CORBA implementation is less of a mainline design issue for the autonomy work. We'll leverage off of whatever is appropriate. We have a possible April simple deployment test for the autonomy for which end of Feb. for testing the communication protocol between the AMC and MVC is important.

• Question:When does the VCS know to expect next behavior from the AMC?
  • First VCS sends AMC a completion message (akin to SUCCESS, FAILURE)
  • VCS will trigger an IDLE state for a fixed time-period (which is parameterized?). An IDLE is a quiescent behavior which will likely shut off the prop and ensure vehicle is not making any changes.
  • AMC will recieve VCS status and send off next commmand to VCS.
  • Meanwhile, also, AMC will expect results of behavior execution to be sent by VCS via a state update.
  • Frederic suggested that one can send two following behaviors so AUV knows what is coming up. Problem with that is that these following behaviors might need to retracted. Even if AMC send's earliest end time that doesn't tell VCS what to do next.
• Rich and Rob are working on how to publish state to the AMC; currently designed the core representation which Conor will be able to use to compile into an agent on the AMC side.
  • They are still working on the task to stick a behavior on the command stack.
  • Setpoint (tells vehicle to run with a command heading angle or depth) is the first behavior to be implemented by Rob & Rich.
• Hans has H/w is bolted up in lab.
  • Will build up Red Hat today.
  • Then will put on the Net to be able to see if additional packages needed by a test compilation.
  • Conor will drop EUROPA inside the stack for such a test compile.
• Clarified that the entire mission script is being run from AMC
  • Instead of reading from a Mission Script, VCS will get the tokens corresponding to the script contents from the AMC
• What do we need when AMC initializes? 
  • look at begining of exisiting mission scripts; Frederic should have copies
  • Primarily, depth envelope and mission timer info needs to be sent over to VCS.
• AMC autonomy code (now called 'T-REX' for "Teleo-reactive EXecutive") will migrate to Subversion on Brian's initiative. We'll us same repository as Brian; will need clients for Mac's and Linux.

From the 2/2 meeting notes, the overal scope of what actually needs to be done, corrsponds to the following:

To be done:

• Behavior stack control

1. 1. start
   2. stop
   3. active status
   4. termination mode 

• More parameters to be exposed from the MVC to the AMC

• For initialization, the AMC needs to send over following information over to the VCS:

1. 1. Timer
   2. depth envelope

• Need to implement an IDLE behavior to ensure that VCS doesn't summarily trigger an abort. One other possible behavior is a "Hang out" behavior when coming early to a waypoint
• Initiate hard abort

Slide1.jpg (image/jpeg)