Ahmad Atta
Knowledge Representation 1. Case Representation 2. Case base representation
General definition [Kolodner, 1993] Problem Description Goals Constraints on the goals Problem situation Solution Solutions Reasoning steps Outcome Expected failure Repair strategy
Case(behavior) in Darmok The declarative part A goal Preconditions Alive conditions The procedural part Basic actions sub goal The state of behavior Pending, executing, succeeded, or failed. The state of goal Open, Ready, Waiting.
Case presentation in CBR-BDI Agents P = where : E is the environment (e.g. game state) O the objectives of agent. O’ the results achieved by the plan R the total resources !! R’ the resources consumed by the agent
Case base representation Case base memory in Darmok consists of two elements: 1. Snippet The procedural part of the behavior. 2. Episode e = ( P, G, S, O) where e.P is the snippet, e.G is the goal, e.S is the situation, and e.O is the outcome of applying e.P in e.S to achieve e,G.
Case memory structure 1. Episodic Memory Organization packets 2. The Category & Exemplar Model
Case Retrieval Case Retrieval in Darmok 1. Episode relevance measure ER ( e, S, G) 2. ER( e, S, G) = GS(e.G, G) + (1-0) SS(e.S, S) 3. Retrieve the episodes with maximum relevance RE(p, S, G) = {e1, ……, ek} 4. Define the predicted performance for each snippets
Adaptation Plan Dependency Graph Generation Precondition-success condition matcher (ps-matcher). Removal of unnecessary actions Adaptation for unsatisfied preconditions