Diagnosing a Team of Agents ScalingUp诊断一个团队的代理比例上升.pptVIP

Behavior Modes Meir Kalech Partially Based on slides of Brian Williams, Luca Console and Peter struss Outline Last lecture: Generation of tests/probes Measurement Selection Probabilities of Diagnoses Today’s lecture: Models of correct + faulty behavior Sherlock engine Abductive diagnosis Qualitative models Exploiting models of correct/faulty behavior Initial proposal: using only models of correct behavior They are those that are in strict accordance with the goals (easy to acquire, e.g., from design) But unfortunately they are not always sufficient Need of exploiting also fault models of some form predictive models [Struss, Dressler, 89] – GDE+[de Kleer, Williams 89] - SHERLOCK “weak” models of physical impossibility [Friedrich et al. 90] behavioral models [Console, Torasso, 91] Diagnosis With Only the Unknown Inverter(i): G(i): Out(i) = not(In(i)) U(i): Diagnosis With Only the Known Inverter(i): G(i): Out(i) = not(In(i)) S1(i): Out(i) = 1 S0(i): Out(i) = 0 Solution: Diagnosis as Estimating Behavior Modes Inverter(i): G(i): Out(i) = not(In(i)) S1(i): Out(i) = 1 S0(i): Out(i) = 0 U(i): Measurement motivation to use Behavior modes Knowledge of failure modes is important to decide what measurement to make next. If all faults were equally likely, measuring X or Y provides equal information. Suppose: Inverters A and B almost always fail by stuck-at-1. Inverter C almost always fails by stuck at-0. It is unlikely that inverter A is failing. The likely failures of inverters B and C are consistent with the symptom Behavior Modes Diagnoses as Assignments of Fault Modes Yet Another Simple Example Fault Localization for the Simple Example - Conflicts 1 and 2 Fault Localization for the Simple Example - Conflicts 3 and 4 Fault Localization for the Simple Example - Hitting Sets What Makes Most of the Fault Localizations Implausible? Fault Models - “Physical Negation” Outline Last lecture: Generation of tests/probes Measurement Selection Probabilities of Diagnoses Today’