Present: Rich, Rob, Hans, Conor and Kanna

•  Rich has an agent class created in the vehicle tree which listens for socket connections for commands  and can send the vehicle state even if there is no listener on client side

• • Send in a string as unencoded text
  • Per Conor, Frederic has implemented a socket on client side; in theory then one can test out the client/server mode of operation of the agents
  • Rich has the code checked into CVS (not SVN); we need to agree on data structure for sending state information (Rob has some test code that needs to be checked into CVS)
  • Rich also has a simple client stub just for testing
  • Has added simple functionality to add some a behavior token in the middle of the stack data structure; note the stack is not a linked list but a dynamic array
• Conor articulated the naming of the entire system by calling it the Teleo-Reactive EXecutive (T-REX)
  • there is now a new Subversion (SVN) project by that name; all new code for the agent will be in SVN
  • Conor has implemented a mechanism to do planer path planning given a set of nodes and links; links are associated with science goal formulations while nodes predomininantly are akin to waypoints used for changing bearing of the vehicle. Such a network restricts spatially what the vehicle can do for a mission.
  • To encapsulate how the behaviors work, we will need to model the interactions of the various behaviors inside the planner domain model. Instead of having a hierarchy of behaviors in a library, we're going to allow all behaviors to be accessible to the VCS as long as they're are in the domain model.
  • One issue is what is needed for cleaning up when a behavior is killed on the stack; e.g if the waypoint behavior is killed while it is executing, then the vehicle will continue to go in the direction the bearing was set. However this same behavior is used to determine when a waypoint is reached; killing it will imply we need a mechanism to deal with such goal achievement, without which vehicle will keep going and not terminate at the waypoint (for instance).
  • Another issue identified was that to manage latency, we will need to have Gulper arming and firing  to be coordinated so that latency is minimized. Need to guarantee that something is on the stack always. The IDLE behavior will mitigate and will possibly need to have some cleanup code.
• Rob indicated that state vector generation was implemented and tested on the Simulator. This is in StatePackage.h
• Hans has the hardware stack in the AUV lab; will have it available and loaded with RH Linux tomorrow (Wed 21st) on the network.
  • Also needs to drill into the glass ball and to be able to stick the stack inside
  • In addition needs to encapsulate a switch to turn the AMC on/off