Robotics

Robotics

1. Agents

   • Autonomous Robots
     • Architecture
       • Rebellion
         • Brittle
           • Computationally Expensive
             • Slow (NP-Hard or worse)
         • Subsumption - Brooks
           • Stateless Behavior-Based Architectures
             • Agent Ignores sensor, does hard-coded thing "always go forward"
             • Act based on sensor, no memory
             • Arbitration
               • Voting
               • Simultaneous
               • Precedence
               • Subsumption
           • Stateful Behavior Based Architectures
             • Need memory if environment doesn't include all clues needed to make decisions
           • Subsumption Architecture
             • Parallel
             • Arbitration
             • Message Passing
             • Inhibit
           • Hybrid - hierarchical with subsumption behaviors
       • 3 tiered
         • The Wrangler
           • Deliberator
             • Executive
               • Control
                 • Effectors
                   • World
                     • Sensors
                       • Feature Selection
     • Agent with real world environment
   • Continuous Behavior
   • FSA Macros
   • Motor Schema Behaviors
   • Multiagent Systems
     • Cooperative
     • Competitive
     • Multiagent Learning
       • Team Learning
         • Learning a solution for multiple agents
       • Collaborative Learning
         • Multiple agents learning solutions
   • Self Contained (no joystick)
   • Environmental Feedback

2. Bayes Filters

   • Kalman
     • Gaussian distribution defines where robot is located
   • Particle
     • Distribution is a random sample of points in your space. Use EC like fitness-based selection to choose parents for next generation of points
   • Tell where the robot is probably located
   • Prior Prob of Robots State
   • Transition Model
   • Sensor Model