[Please remove <h1>]

---

David R. Cheriton (Stanford University) – Distributed Structuring of a Blackboard-based Diagnosis Application

Complex systems can develop complex problems resulting in situations that are difficult to interpret. Significant research has been done some years ago under the label of ‘blackboard systems’ to diagnosis situations by utilizing separate logical processes updating a shared state (the “blackboard”) which consists of sensor data and a representation of multiple possible scenarios describing the situation ranked by ‘confidence’ in each possibility. Previous systems based on this architecture, although shown to diagnose effectively, were implemented in a single address space with extraneous complexity, resulting in a system that was difficult to maintain and update.

In this talk, I describe a distributed blackboard-based diagnosis architecture, explored using the (toy) example domain of vessel surveillance, and with separate processes for each level of knowledge leading to a modular and extensible design, avoiding the ‘arbiter’mechanism required in previous ork. Our system demonstrates basic diagnostic capabilities while exhibiting the fault-tolerance, parallelism and extensibility of a true distributed implementation.

Location: DC 1302
Time: 2:00-2:40pm