Jam agents in a Nutshell Version 65 76i 1
Download 391.89 Kb.
|
precond.jamThis file tests the precondition and effects sections of plans and highlights the functional difference between the precondition and context fields. lisplist.jamThis file tests and demonstrates the string and token-handling primitive actions suggested and originally implemented by Chip McVey at the Johns Hopkins University Applied Physics Lab (JHU-APL). succeedfail.jamThis file is anachronistic and effectively only tests and and or branch constructs. The comments indicate a point in development when these constructs exhibited non-deterministic behavior. goalsemantics.jamThis file demonstrates the difference between achieve and perform plans and goals. Notice the difference in behavior when achieve and perform are switched in the subgoaling actions and the plan’s goal specification. There is also two assert actions in the subgoal that can be uncommented to modify the agent’s behavior. These actions perform “opportunistic accomplishment” of the respective goals when the parent or subgoal are invoked as achievement goals. maintain.jamThis file tests and demonstrates maintain goals. post.jamThis file tests the post and unpost actions. doallany2.jamThis file tests the do_all and do_any constructs in conjunction with other JAM constructs. threads2.jamThis file tests subgoaling from within a parallel construct. Note that the subgoaling works because subgoal2 is achieved in the second branch before the first branch tries to subgoal to subgoal1. Simultaneous parallel subgoaling is not currently supported and will result in one branch of the parallel construct never completing. wait.jamThis file tests the two forms of the wait construct. conclude.jamThis file test plans with conclude specifications (i.e., data-driven) intermixed with standard goal-driven behavior (both achieve and maintain). Note that the goals, plans, and facts are mixed throughout the file. There is nothing new with this, but the previous files do not demonstrate this “feature”. The overall behavior of this example is that the agent will start working on the top-level goal and then notice every 125 cycles that the conclude specification is met and will exhibit some data-driven behavior. The goal-driven plan will be suspended temporarily because the utility of the data-driven plan is higher than the goal-driven one. The maintain goal state will be undone four times before the agent removes that goal and exits. metalevel.jamThis file tests and demonstrates handling of arbitrary Java objects and metalevel reasoning behavior and primitive actions. parse-string.jamThis file tests and demonstrates runtime, dynamic parsing of information into the agent’s World Model and Goal List. sockets-client.jamsockets-client2.jamThis file tests and demonstrates use of the socket-based communication primitives. Each of these files are used in conjunction with the corresponding sockets-server file below (i.e., sockets-client.jam with sockets-server.jam, and sockets-client2.jam with sockets-server2.jam). As indicated by the names, these files are used by a JAM agent connecting as a client to another agent. sockets-server.jamsockets-server2.jamThis file tests and demonstrates use of the socket-based communication primitives. Each of these files are used in conjunction with the corresponding sockets-server file above (i.e., sockets-client.jam with sockets-server.jam, and sockets-client2.jam with sockets-server2.jam). As indicated by the names, these files are used by a JAM agent acting as a server for connecting to another agent.
The JAM agent grammar starts with the "ParseJamAgentDefinition" symbol. NON-TERMINALS
|