T. Sugawara, "Reusing past plans in distributed planning," Proc. of the Int. Conf. on Multi-Agent Systems (ICMAS95), 360 - 367, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....both learning and planning are jointly and distributedly realized by multiple agents. In the area of multiagent systems a lot of work is available on both activity coordination through joint learning (e.g. 1, 12, 15, 16, 18, 19, 26, 27] and activity coordination through joint planning (e.g. [2, 3, 6, 8, 10, 11, 17, 20]) However, there are only very little approaches that combine joint learning and joint planning. There are two exceptions that are related to the JPJL algorithm. The first is the work by Sugawara and Lesser described in e.g. 21, 22] The basic idea behind this approach is to enable agents to ....

T. Sugawara, `Reusing past plans in distributed planning', in Proceedings of the First International Conference on Multi-Agent Systems (ICMAS-95), pp. 360--367, (1995).

....coordinate their individual activities such that the resulting overall activity patterns are useful with respect to attaining their goals and accomplishing their tasks. There are two contrary main approaches that aim at answering this question: one concentrates on plan based coordination (see e.g. [3, 4, 5, 9, 12, 16, 25, 30, 17]) and the other on reactive coordination (see e.g. 1, 8, 10, 19, 22, 24, 27, 29] and also [6] The basic idea behind plan based coordination is that the agents jointly generate hypothetical activity sequences on the basis of their world model in order to nd out in advance (i.e. before acting ....

.... the start state into the goal state (successful sequence) or into a non goal state (unsuccessful sequence) The parameter setting was as follows: 0:2, 0:9, and R = 1000 (SAS1 60 100 140 180 220 260 300 340 380 20 200 400 600 800 1000 avg reward RANDOM MDYNAQ[10] MDYNAQ[0] episodes MDYNAQ[30] Figure 4: Experimental results for SAS1. 60 100 140 180 220 260 300 340 380 20 200 400 600 800 1000 avg reward MDYNAQ[10] RANDOM MDYNAQ[0] episodes MDYNAQ[30] Figure 5: Experimental results for SAS2. in state 17) R = 500=1000 (SAS2 in states 16 20) The initial Q values were all zero. Each ....

[Article contains additional citation context not shown here]

T. Sugawara. Reusing past plans in distributed planning. In Proceedings of the First International Conference on Multi-Agent Systems (ICMAS-95), pages 360-367, 1995.

....system performance. In the area of multiagent systems a lot of work is available on both activity coordination through joint learning (e.g. CB98, Mat97, OF97, Par97, SV96, SS96, Wei93a, Wei93b] and activity coordination through joint planning (e.g. DL92, DKK97, DL91, Geo83, HD96, Kab95, SH96, Sug95, von92] However, there are only very little approaches that combine joint learning and joint planning. There are two exceptions that are related to the JPJL algorithm. The rst is the work by Sugawara and Lesser described in e.g. SL93, SL98] The basic idea behind this approach is to enable ....

T. Sugawara. Reusing past plans in distributed planning. In Proceedings of the First International Conference on Multi-Agent Systems (ICMAS95) , pages 360-367, 1995.

....learning, but the reader is directed to (Garland Alterman 1996) for more technical details on operator probability trees. Case based reasoning has been previously demonstrated to be an effective technique for reducing CPU usage during planning in a standard distributed AI planner in (Sugawara 1995). In our domain though, CPU usage is an inappropriate measure of community performance because we are primarily interested in improving the community s run time behavior, not the speed at which they plan (or retrieve plans from memory) However, since primitive actions and communication are ....

Sugawara, T. 1995. Reusing past plans in distributed planning. In Proc. First International Conference on Multiagent Systems, 360--367.

....and the plan coordinate execute cycle is repeated. Obviously, if this E 406 happens frequently, a substantial expenditure of effort for planning and E 376 coordination can result. Sometimes, strategies such as repairing the previous plans, E 433 or accessing a library of reusable plans [Sugawara 1995] can reduce the effort to make E 434 it managable. E 12 Significant overhead can of course be saved if a plan deviation can be addressed E 386 locally rather than having to require coordination. For example, rather than E 400 coordinating sequences of actions, the agents might coordinate their ....

Toshiharu Sugawara. Reusing past plans in distributed planning.<E-403> Proceedings of the First International Conf. on Multi-Agent Systems (ICMAS-95),<E-379> pages 360-367, June 1995.<E-108>

....optimality of the composite case. However, the notion of globally situating local cases is not a part of this work. In addition, though some of the ideas have been tested empirically, the test domains are of much lesser complexity than the kind of domains dealt with in this thesis. Sugawara[Sugawara, 1995] presents a hierarchical planner that can use past plans to minimize messages at planning time. It is based on Corkill s[Corkill, 1979] approach to distributing the hierarchical NOAH[Sacerdoti, 1977] planning system over multiple agents. Coordinating consists of avoiding undoing each other s work ....

....executing action A, it communicates a wait synchronization message to the other agent and inserts a local signal message after action A. The other agent inserts a wait message before action B preventing the execution of the action until the corresponding signal message is received. Sugawara[Sugawara, 1995] extends this 24 mechanism to let an agent store its plan along with the synchronization sequence for a given problem solving run. The plan is indexed using the initial and goal states of the problem instance. When a similar problem instance occurs later, the past plan is reused and the need ....

[Article contains additional citation context not shown here]

Sugawara, T. Reusing past plans in distributed planning. In Proceedings of the First International Conference on Multi-Agent Systems, pages 360--367, San Francisco, CA, June 1995. AAAI Press.

No context found.

T. Sugawara, "Reusing past plans in distributed planning," Proc. of the Int. Conf. on Multi-Agent Systems (ICMAS95), 360 - 367, 1995.

No context found.

T. Sugawara. Reusing past plans in distributed planning. In Proceedings of the First International Conference on Multi-Agent Systems (ICMAS-95), pages 360--367, 1995.

No context found.

T. Sugawara, `Reusing past plans in distributed planning', in Proceedings of the First International Conference on Multi-Agent Systems (ICMAS-95), pp. 360--367, (1995).

No context found.

Toshiharu Sugawara. Reusing past plans in distributed planning. In Proc. First Intl. Conference on Multiagent Systems, pages 360--367, 1995.

No context found.

Sugawara, T. (1995). Reusing past plans in distributed planning. In Proc. First International Conference on Multiagent Systems, pages 360--367.

No context found.

Sugawara, T. (1995). Reusing past plans in distributed planning. In Proc. First International Conference on Multiagent Systems, pages 360--367.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC