K. Toyama and G. Hager. If at first you don't succeed... In Proc. AAAI, pages 3--9, Providence, RI, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....multiple agent models. We have implemented SAM in a complex, realistic multi agent domain, and provide detailed empirical results assessing its benefits. Introduction Attaining robustness in face of uncertainty in complex, dynamic environments is a key challenge for intelligent agents (Toyama and Hager 1997). This problem is exacerbated in complex multi agent environments due to the added requirements for communication and coordination. Example domains include virtual environments for training (Tambe et al. 1995) robotic soccer (Kitano et al. 95) potential multi robotic space missions, etc. The ....

....opportunities for 1 Copyright 1998, American Association for Artificial Intelligence (www.aaai.org) All rights reserved. failure, and must autonomously monitor and detect failures in their run time behavior, then diagnose and recover from them, i.e. agents must display post failure robustness (Toyama and Hager 1997). Previous approaches to monitoring and diagnosis (e.g. Doyle et al. 1986, Williams and Nayak 1996) have often focused on a single agent that utilizes designer supplied information, either in the form of explicit executionmonitoring conditions, or a model of the agent itself. This information ....

Toyama, K.; and Hager, G. D. 1997. If at First You Don't Succeed..., in Proceedings of the Fourteenth National Conference on Artificial Intelligence (AAAI-97). pp. 3-9.

....the ability of an agent modeling algorithm to represent uncertainty and the applications, and highlight key lessons learned for real world applications. Introduction A key challenge in using intelligent systems in real world environments is the attainment of robustness in face of uncertainty (Toyama and Hager 1997). The explosion of state space complexity completely inhibits the ability of any designer, human or machine, to specify in advance the correct response in each possible state (Atkins et al. 1997) In addition, an agent in such environments cannot completely and correctly sense their environment at ....

....teams to perform multiple missions over the duration of a major simulation exercise requires a very high degree of robustness from participating agents and agent teams. To attack this problem, AI techniques have attempted to provide agents with the capability to demonstrate post failure robustness (Toyama and Hager 1997), which allows the agents to autonomously detect, diagnose, and recover from failures as they occur. Previous approaches (e.g. Doyle et al. 1986, Williams and Nayak 1996) have attempted to improve the robustness of agents performance by specifying constraints which essentially allow an agent to ....

[Article contains additional citation context not shown here]

Toyama, K.; and Hager, G. D. 1997. If at First You Don't Succeed..., in Proceedings of the Fourteenth National Conference on Artificial Intelligence (AAAI-97). pp. 3-9.

....to the system designer (W. Hamscher 1992) Our goal was to use a technique that offered great flexibility in the information it could use and the diagnoses it could generate, without sacrificing subject scope or domain independence. Recent work in the field of diagnosis (Kaminka Tambe 1998; Toyama Hager 1997) has shown that viable new technologies are still being developed. It is not clear, however, that any single diagnostic technique is suitable for the entire range of faults exhibited by multi agent systems. It was therefore desirable to use a system or framework capable of incorporating different ....

....systems, ranging from fully explicit, verbose communication to well known assumptions or implicit agreements. Clearly, it is more efficient to reduce inter agent communication if possible, but how can an agent know when it is safe to do so One method, similar to a technique described in (Toyama Hager 1997), makes use of a persistent diagnostic process to monitor tested changes. In this example, the UnecessaryRsrcCoordination node begins its work by monitoring the coordination which takes places over the system s resources. If it detects that requests for a particular resource are always being ....

Toyama, K., and Hager, G. D. 1997. If at first you don't succeed... In in Proceedings of the 14th National Conference on Artificial Intelligence (AAAI97) . AAAI.

....to be practical, face tracking must be robust. In particular, because users cannot be tracked perfectly all of the time (for example, when they leave the field of view) tracking systems must be post failure robust and capable of recovering from any visual disturbance that disrupts normal tracking [19]. 2.1 Incremental Focus of Attention Reliable face tracking is accomplished by implementing a system based on Incremental Focus of Attention (IFA) an architecture for incorporating different tracking algorithms into a robust real time tracking system [18, 20] If designed for the face tracking ....

K. Toyama and G. Hager. If at first you don't succeed... In Proc. AAAI, pages 3--9, Providence, RI, 1997.

....problem is, therefore, a vision based tracking subproblem the problem of coping with a complex environment. 3 Previous Work in Robust Tracking The existing literature on robust tracking can be broadly categorized into research that contributes to either ante failure or post failure robustness [42]. Ante failure robust systems seek to avoid tracking failure altogether through specialized algorithms that anticipate visual disturbances and attempt to track despite them. Post failure systems, on the other hand, accept the inevitability of mistracking and are designed to recover from failure ....

K. Toyama and G. D. Hager. If at first you don't succeed... In Proc. Nat'l Conf. on AI, Providence, RI, 1997.

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