AuRA: Principles and Practice in Review

1
AuRA Principles and Practice in Review

• Paper by
• Ronald C. Arkin and Tucker Balch
• Present By
• Jirakhom Ruttanavakul

2
Introduction

• AuRA Autonomous Robot Architecture
• AuRA was developed in mid-1980s (as a hybrid
  approach to robotic navigation)
• Arose from
• A deliberative or hierarchical planner
• A reactive controller

3
Guideline

• The structure of AuRA
• The strengths of AuRA
• The origin of AuRA Theory
• The example of AuRA-Based System
• Conclusion

4
Structure of AuRA
Learning
User Input
REPRESENTATION
Plan Recognition User Profile
User Intentions
Mission Planner
Hierarchical Component
Spatial Reasoner
Spatial Learning
Spatial Goals
Plan Sequencer
Mission Alterations
Opportunism
Schema Controller
Reactive Compnenet
On-line Adaptation
Teleautonomy
Motor
Perceptual
Actuation
Sensing
5
Structure of AuRA (Cont.)
Mission Planner
Spatial Reasoner
Plan Sequencer

• Mission Planner Establishing high-level goals
  and constraints within which it must operate.
• Spatial Reasoner Using the knowledge in
  long-term memory to construct a sequence of path
  legs.
• Plan Sequencer Translating each path legs into a
  set of motor behaviors (schemas)
• The schemas will be sent to the robot.
• The deliberative system will stop and reactive
  system will start.

6
Structure of AuRA (Cont.)
Schema Controller
Motor
Perceptual

• Schema Manage Controlling and Monitoring the
  behavioral processes at run-time.
• Motor Schema associated with Perceptual schema
  Providing the stimulus required for that
  particular behavior.
• Homeostatic Control System Maintaining balance
  and system equilibrium.
• Hierarchical Component will be reactivated, only
  if a failure is detected (lack of progress,
  velocity of zero, and timeout)

7
The Strengths of AuRA

• Modularity Components can be replaced with
  others in straightforward manner
• Flexibility It provides for introducing
  adaptation and learning methods.
• Generalizability
• Hybridization Gats Atlantis Architecture, 3T

8
The Origin of AuRA Theory

• Aura influenced by a wide range of ethological,
  neuroscientific, and psychological study
• Schema Theory a theory of intelligence which
  represents motor and perceptual control at a
  level of abstraction higher than that of neural
  networks. The AuRA employs at the reactive
  control level, encoded using an analog of the
  potential fields method
• Justification for hybridization of reactive and
  deliberative control found in studies by
  psychologists
• Homeostatic Control System developed using
  models of the mammalian endocrine system as
  inspiration.

9
The Example of AuRA-Based System

• Trash-Collecting Robots Built by a Group of
  Georgia Tech Student in 1994
• Objective Searching for trash, Picking it up,
  Carrying it to the wastebaskets
• The trash consisting of Styrofoam coffee cups,
  wads of paper, and soda cans.
• The environment consisting of obstacles such as
  tables and chairs.

10
Robots Hardware and Sensing

• Power System Computer Equipment
• Sensors including bumper switches for collision
  detection
• Color Video Camera The key factor of the
  robots success in their task
• A custom-built gripper attatched to the front
  of the robots
• Infrared Sensor mounted in the gripper

11
Low-level Behaviors for the Robots

• The lowest level motor schemas
• Schema Controller instantiates and runs schemas
  as directed by the Plan Sequencer.
• A set of schemas is active at a time
• Each motor schemas Computing a vector which
  indicates a desired direction of motion.
• The vector is combine to generate the overall
  movement vector
• The overall movement sent to the robots
  actuator

12
The Example of Schemas

• Detect-red-blob using vision to find the
  location of the goal (red is trash, blue for
  wastebaskets, green for robots).
• Detect-obstacles using bumper switches to
  detects and tracks obstacles
• Move-to-goal generating a vector towards the
  goal found by detect-red-blob
• Avoid-static-obstacles generating a vector away
  from any detected obstacles
• Detect-IR-beam-broken used as the trigger to
  close the gripper around the object

13
A Plan for Robots

• The plan, coded by humans, is a sequence of
  behavioral assemblages and perceptual triggers
  which causes the transition between them,
  expressed as a Finite State Acceptor (FSA)
• States identified with circles
• Perceptual Triggers directed arcs between
  states
• When the condition on arcs is met, the state will
  be changed

14
(No Transcript)
15
Cooperation in Robots

• No communication devices with the robots
• Simply paint the robots to green color.
• The robot move away from the green in
  wander-for-trash state

16
Conclusions

• The AuRA is a hybrid architecture which combine
  deliberative planner, based on traditional AI
  techniques and reactive controller, based on
  schema theory.

17

• Thank You
• Questions Comments