Deliberative Systems 3 phase model: –Sense –Plan –Act Advantages: –can plan –Can learn Disadvantages: –Needs world model –Searching and planning are slow –World model gets outdated
Feedback Control React to the sensor changes Feedback control == self-regulation Q: What type of control system is it? Feedback types: –Positive –Negative
- and + Feedback Negative feedback: –Regulates the state/output –Examples: Thermostat, bodies, … Positive feedback: –Amplifies the state/output –Examples: Stock market The first use: ancient Greek water system Re-invented in the Renaissance for ovens
W. Grey Walter’s Tortoise 1953 Machina Speculatrix Sensors –1 photocell, –1 bump sensor 2 motors Reactive control
W. Grey Walter’s Tortoise Behaviors: seeking light, head toward weak light, back away from bright light, turn and push (obstacle avoidance), recharge battery. Basis for creating adaptive behavior-based
Turtle Principles Parsimony: simple is better –e.g., clever recharging strategy Exploration/speculation: keeps moving –except when charging Attraction (positive tropism): –motivation to approach light Aversion (negative tropism): –motivation to avoid obstacles, slopes Discernment: ability to distinguish and make choices –productive or unproductive behavior, adaptation Ducking
Tortoise behavior A path: a candle on top of the shell
Tortoise behavior Two turtles: Like dancing
Question How does it do the charging? –Note: When the battery is low, it goes for the light.
Braitenberg Vehicles Valentino Braitenberg –early 1980s Extended Walter’s mode Based on analog circuits Direct connections between light sensors and motors Complex behaviors from very simple mechanisms
Braitenberg Vehicles Complex behaviors from very simple mechanisms
Braitenberg Vehicles By varying the connections and their strengths, numerous behaviors result, e.g.: –"fear/cowardice" - flees light –"aggression" - charges into light –"love" - following/hugging –many others, up to memory and learning! Reactive control Later implemented on real robots Check: Bots order Styrofoam cubes (16 min 30 sec) –Tokyo Lecture 3 time 24:30-41:00
Brief History 1750: Swiss craftsman create automatons with clockwork to play tunes 1917: Word Robot appeard in Karel Capek’s play 1938: Issac Asimov wrote a novel about robots 1958: Unimation (Universal Automation) co started making die-casting robots for GM 1960: Machine vision studies started 1966: First painting robot installed in Byrne, Norway. 1966: U.S.A.’s robotic spacecraft lands on moon. 1978: First PUMA (Programmable Universal Assembly) robot developed by Unimation. 1979: Japan introduces the SCARA (Selective Compliance Assembly Robot Arm).
Early Artificial Intelligence "Born" in 1955 at Dartmouth "Intelligent machine" would use internal models to search for solutions and then try them out (M. Minsky) => deliberative model! Planning became the tradition Explicit symbolic representations Hierarchical system organization Sequential execution
Artificial Intelligence Early AI had a strong impact on early robotics Focused on knowledge, internal models, and reasoning/planning Eventually (1980s) robotics developed more appropriate approaches => behavior-based and hybrid control AI itself has also evolved... Early robots used deliberative control Intelligence through construction (5 min 20 sec) –Tokyo Lecture 2 time 27:40-33:00