---

Patrick Henry Winston
Professor of Artificial Intelligence and Computer Science at the Massachusetts Institute of Technology
Director of the Artificial Intelligence Laboratory from 1972 to 1997
Chairman and co-founder of Ascent Technology, Inc.
Personal Website 
Email: Этот адрес электронной почты защищён от спам-ботов. У вас должен быть включен JavaScript для просмотра.

 

Общие сведения

Архитектура CHIP разработана в 2006 г. исследовательской группой под руководством Говарда Шробе (Howard E. Shrobe), Патрика Уинстона (Patrick H. Winston), в Лаборатории искусственного интеллекта (CSAIL)  Массачусетского технологического института, Кембридж, Массачусетс, США.

---

The Comprehensive Human Intelligence Project (CHIP) architecture

The Comprehensive Human Intelligence Project (CHIP) architecture, is an evolving framework for building a system that attempts to encompass the full range and magic of human cognition.

The Comprehensive Human Intelligence and Performance (CHIP) architecture [Shrobe, et al., 2006] is a multi-level hybrid architecture that provides a good illustration of a three-layer design that incorporates reactive, deliberative, and reflective control. At each successively higher layer, the control processes simultaneously become slower and yet capable of handling more complex problems. In addition to the three layers of control, CHIP also uses the sequential sense-plan-act (SPA) architectural schema common to robotic control systems.

The CHIP architecture can be represented generally using a 3x3 grid as shown in Figure 1.1.

 

Figure 1.1: The CHIP Architecture: the systems of components intimately interact with one another,
forming tightly integrated loops through the various layers. [Shr, Win, 2006]

---

CHIP is one of many architectures that trifurcate control in this way. In each of these architectures, the lowest level of  control performs reactive or reflexive responses (e.g., an emergency stop before a collision).

The next level of control varies slightly from architecture to architecture but is generally responsible for performing simple deliberation or routine skills. This primarily involves control just beyond that afforded by simple reactions (e.g., following a path or grasping a familiar object) [Ortony, et al., 2004].

The final layer performs the most complex form of control, which is often some form of complex deliberation, reflection, or meta-management.

This level of control is typically capable of generating and analyzing plans, arbitrating among abstract response options, biasing parameters in the lower control levels, and performing post hoc evaluations.

---