Subsumption Architecture and Nouvelle AI Arpit Maheshwari Nihit
Subsumption Architecture and Nouvelle AI Arpit Maheshwari Nihit Gupta Saransh Gupta Swapnil Srivastava
Seminar Roadmap 1. PSS and Knowledge Representation 1. 1 Basic Idea 1. 2 Problems with Abstraction 2. Nouvelle AI 2. 1 Framework 2. 2 Decomposition by activity 2. 3 Differences with Classical AI 2. 4 Methodology in practice: Subsumption Architecture
Roadmap (Contd. . ) 2. 5 Challenges 3. Summary Comparision: Classical vs Nouvelle AI
1. PSS and Knowledge Representation • A physical symbol system consists of a set of entities, called symbols, which are physical patterns that can occur as components of another type of entity called an expression (or symbol structure) • A physical symbol system is a machine that produces through time an evolving collection of symbol structures
PSS Hypothesis • A physical symbol system has the necessary and sufficient means for general intelligent action. Allen Newell and Herbert Simon, 1975
Classical framework Perception Model Plan and Act
Problems with Abstraction • Intelligence = Abstraction + Reasoning (Logically) • The efforts at AI are not truly intelligent (Why? ) • Claim: An abstraction would never be as informed as the object itself e. g. chair
Problems with Abstraction(contd. . ) Human: Sensing Intelligence Machine: Sensing Abstraction Reasoning Example- Chess playing
2. Nouvelle AI • Also called Behavior-based AI • It is extremely popular in robotics • It allows the successful creation of real-time dynamic systems that can run in complex environments.
Framework • Concept of a “Creature” – an engineering methodology • Incremental Intelligence • Testing in Real World “The world is the best model of itself” • Intelligence stems from a tight coupling between sensing and actuation (No knowledge representation)
Evolution: A motivation Expert Systems insects single-celled life 3. 5 billion years ago Brooks’ conclusion: Complex behavior, knowledge, and reason are all relatively simple once the basics of survival - moving around, sensing the environment, and maintaining life - are acquired. 550 million years ago present day
Decomposition by Activity • Layer: An activity-producing system • Each activity connects sensing to action directly • Advantage- A clear incremental path for simple to complex systems. Easy to add behaviors
What is different? • No specific output of perceptions • No Central System • Representation got rid off Example: Eye sensing
Society of mind • Proposed by Minsky • Nouvelle AI seems to draw inspiration from this concept
Methodology in practice • Subsumption Architecture Developed by Rodney Brooks for robot control in 1986
Earlier approach-Function modules Sensors Actuators
Layered Architecture The Subsumption Architecture is: • A layering methodology for robot control systems • A parallel and distributed method for connecting sensors and actuators in robots
An example: A mobile robot Layer 5: Identify objects Layer 4: Monitor changes Layer 3: Build maps Layer 2: Explore Layer 1: Wander aimlessly Layer 0: Avoid hitting objects
Merits • Multiple Goals • 2 -fold Robustness • Additivity
Structure of Layers • Each layer is made up of connected, simple processors: Augmented Finite State Machines
Layers (contd. . ) • The most important aspect of these FSMs – Outputs are simple functions of inputs and local variables – Inputs can be suppressed and outputs can be inhibited • This function allows higher levels to subsume the function of lower levels • Lower, therefore, still function as they would without the higher levels
Nouvelle AI Different from • Connectionism, Neural networks • Production rules system
Challenges • Maximum number of layers? • How complex can the behavior be that are captured without central representation? • Can higher-level functions such as learning occur?
Summary • • Classical AI Make a detailed static plan in advance Representation-based Simplified-world Central and peripheral systems Nouvelle AI React directly to the world No central representation Real world No such distinction
References 1. R. A Brooks (1991). "Intelligence Without Representation", Artificial Intelligence 47 (1991) 139 -159 2. Brooks, “A Robust Layered Control System for a Mobile Robot”, Robotics and Automation, IEEE Journal of; Mar 1986, pp. 14 – 23, vol. 2, issue 1
- Slides: 25