Revolution in Robotics and Intelligent Control by yurtgc548


a new paradigm

Dr. Reuven Granot
Courtesy of the Center for Robot-Assisted Search and Rescue.
BBC News   Monday, 17 September, 2001, 15:20 GMT

Robots aid New York rescue workers

       Smoldering ruins: Robots can go where people
  • Robots are being used to search for victims amid the rubble of the
  twin towers of the World Trade Center in New York.
  • Three small experimental robots are being lowered into gaps
  between collapsed buildings to assist in the recovery of bodies.
  • They are each about the size of a shoebox and are operated by hand.
                   National Institute for Search and Rescue
              Center for Robot-Assisted Search and Rescue

Summary statistics of the first 11 days of rescue and recovery search
Arrival at Stewart Air Field: Tuesday, Sept 11, 17:30
Arrival at JAAVITS center: Wednesday, Sept 12, 01:00
Arrival at Ground Zero: Wednesday, Sept 12, 06:00

Victims found: 3-5 – by MicroTracs and MicroVGTV
Voids searched: 2 (1 declared safe by rescuers & a victim found upon entry, 1 declared unsafe) - by
MicroTracs and Solem

Buildings searched: 3 from 5 viewpoints-by UrBot, PackBot, Talon
Number of robots on site: 17 (though the number varied)
Robots deployed at Ground Zero: 8
Personnel on-site: in addition to Blitch, Foster-Miller (3), iRobot (8), SPAWAR (3), USF (4)
Lost robots: 1 – Solem
Damaged robots: 2 – MicroTracs and MicroVGTV, both repaired on-site

Number of excursions to Ground Zero: 11- 5 times actually inside the rubble pile on WTC 2, Marriott,
and surrounding areas, the rest on-station with task forces
Size of robots: MicroTracs and MicroVGTV are size of a shoebox and use tethers, others are size of small suitcase and
can be operated through wireless.
Sewer Insert
A void searched by Solem


                           Courtesy of the Center for Robot-Assisted Search and Rescue.
A building searched by   UrBot


                                 Courtesy of the Center for Robot-Assisted Search and Rescue.
A building searched by   PackBot


                               Courtesy of the Center for Robot-Assisted Search and Rescue.
       Joint development of NASA/JPL, iRobot, CMU and USC
                     under DARPA/ATO contract

Purpose: mobile reconnaissance in city, police and rescue personnel.
     The platforms were developed under DARPA contracts
                   Tactical Mobile Robotics project
Concept                                 Approach:
Penetrate denied areas and project
operational influence in ways that      • Integrate sensors, locomotion,
humans cannot by using reliable semi-   power, communications, and
autonomous robotic platforms.           sufficient smarts
                                        • on a compact, man-portable
                                        • to provide a semi-autonomous
                                        • capable of serving as an extension
                                        of the human soldier.

  Top Technical Challenges
      Robotic mobility in cluttered and complex terrain
      Machine perception for obstacle negotiation        PerceptOR
      Autonomous operation and fault recovery
  Examples of Robots                ISR Urban robot

• ISR Urban “Cowboy”
• Inuktun MicroVGTV and Inuktun
micro-track pipe crawler

• SAIC Subot
 Inuktun microVGTV and micro-trak

                                    SAIC Subot
Urban Search and Rescue (USAR) Robot Competitions
    Robots made by teams from around the world went on
    simulated search-and-rescue missions in Seattle on August 4,
    2001 at RoboCup Rescue 2001.
2001+ RoboCup USAR                  NIST Standard Test Course for USAR
Physical Agents League              prototype
   – May be untethered (no
   chain), but human in the loop
   – Robots (and their supporting
   workstations) are a “member”
   of the technical rescue team
  • Mobility
  • Perception
  • Decision making
      – Autonomously
      – Intelligently
            Problems with Mobility
Most robots tested were unable to cope with rubble
    -But weren’t designed to

                                                          June 2, 2000 SRDR Miami Beach:
                                                          Types of clutter
                                                                           WTC in NY 11 Sept 2001.

                                            Courtesy of the Center for Robot-Assisted Search and Rescue.
       Problems with Mobility
Rubble at WTC NY Sept 2001

                             Courtesy of the Center for Robot-Assisted Search and Rescue.
                Problems with Mobility
Most robots tested were unable to cope with rubble
     -But weren’t designed to

Urban able to cope with rubble, but sensitive to
tracks containing plastic bags, sheets, throw rugs,
draperies, etc

                                                            June 2, 2000 SRDR Miami Beach:
                                                            Types of clutter
                                                                             WTC in NY 11 Sept 2001.

  June 2, 2000 SRDR Miami Beach: plastic
  bag caused Cowboy to throw trackpp
                                              Courtesy of the Center for Robot-Assisted Search and Rescue.
  Problems with Perception
Example: while urbot is climbing stairs
   • WTC NY Sept 2001

                               Courtesy of the Center for Robot-Assisted Search and Rescue.
DARPA                                                                       PerceptOR
                  Perception for Off-Road Robotics
                          Expected Trends
                   Spectrum of Robotic Autonomy
       Manned   operation              Semi-autonomous                   Autonomous

       Level of human        Today
         interaction                   Focus of PerceptOR – applied to                Term
                                            autonomous mobility                       Goal

       Manned                          Semi-autonomous                   Autonomous

         PerceptOR seeks to define the reduction of human
         interaction based on improvements in autonomous mobility.
 Some relevant technologies are already available and
 used in newly developed intelligent toys and humanoids.
AIBO Entertainment robot        Humanoids from Honda
from Sony
                                            • RoboCupJunior is a project-oriented educational initiative.
                                            • It is designed to introduce RoboCup to primary and
By 2050, develop a team of fully            secondary school children.
autonomous humanoid robots
that can win against the human
world champion team in soccer.                                         RoboCupSoccer is divided into
                                                                       the following leagues:
                                                                       • Simulation league

RoboCupRescue Search                                                   • Small-size robot league
and Rescue for Large                                                   • Middle-size robot league
Scale Disasters:                                                       • Four-legged robot league
• simulation project                                                   • Humanoid league
• physical robots project.

                       RoboCup-2002 Fukuoka / Busan June 19th - June 25th, 2002
                       The Sixth Robot World Cup Soccer Games and Conferences,
                       Official Site:
A paradigm is a philosophy or set of assumptions and
              techniques, which characterize an
              approach to a class of problems.
 A machine can be distantly operated by:
     • continuous control: the HO is responsible to continuously
     supply the robot all the needed control commands.
     • a coherent cooperation between man and machine, which
     is known to be a hard task.

  A telerobot is a robot that determines its actions
  based on some combination of human input and
  autonomous control.

Telerobotics is a form of Supervised Autonomous Control.
              The spectrum of control modes.
                                      A telerobot can use:               • traded control:
                                                                           control is or at
                                                                           operator or at the
                                                                           autonomous sub-
                                                                         • shared control: the
                                                                           instructions given by
                                                                           HO and by the robot
                                                                           are combined.
                                                                         • strict supervisory
                                                                           control: the HO
                                                                           instructs the robot,
                                                                           then observes its
                                                                           autonomous actions.

Solid line= major loops are closed through computer, minor loops through human.
              The need for
         Collaborative Research

•   The task is hard to achieve
•   Needs expertise in many disciplines
•   Needs state-of-the-art components
•   Fast advance/ quick changes in the required
          Collaborative Research
The incubation phase has to be done on shared
research test bed that is:
– supplied and supported through the Internet by
  partner research groups,
     • as opposed to developed products, which are limited
       by the negotiated financial, technical and market
       oriented constrains.
–   agreed on paradigm and architecture
–   state-of-the-art components
–   highly reliable environment (quality)
–   developed on agreed standards
    Robot Architecture Major Classes/Categories
    A control architecture provides a set of principles for
    organizing a control system.
        – It provides structure and constraints which aid the designer
        in producing a well-behaved controller.

     Intuitively, this means that there are infinitely many ways to structure
     a robot program, but they all fall into one of major classes
     /categories of control:

   deliberative           look-ahead: think/plan, then act
   reactive               no look-ahead: react
3) behavior-based         distribute thinking over acting
   hybrid                 combine 1+2, think slowly, react quickly
     Real time Control Systems
   A Reference Model Architecture
       For Teleoperation and Telerobotic applications

• RCS allows for
  shared control as
  well as for strict
  control, in which
  HO can directly
  instruct each
  intelligent node.
RCS adopts a
 • hierarchical architecture
 • multilevel, each level deals with different resolution in
      space (range)
      time
The Behavior Generator Module in RCS

                        Agents act on behalf of
                        their supervising agent
                           • assign jobs
                           • plan
                           • select and
                           coordinate plans
                           • execute
  What are Behaviors?
• typically have the following properties:
    – are feedback controllers but extended in time
    – achieve specific tasks/goals
    – can directly connect sensors and effectors
• When assembled into distributed representations, behaviors can be
  used to look ahead but at a time-scale comparable with the rest of the
  behavior-based system.

   An individual behavior is a stimulus/ response pair for a
   given environmental setting that is modulated by attention
   and determined by intention.
     prioritizes tasks and focuses sensory resources
     determines which set of behaviors should be active
            based on the robotic agent’s goals and objectives.
                                            Search and planning takes too long

                         Deliberative Systems
     Based on the Sense  Plan  Act model
     Inherently sequential                         Problem 1: Time Scale
                                                   "too much information"
     Planning requires search
                                                    Generating a plan is slow.
     Search requires a world model                Problem 2: Space
      The representation must be constantly        Generating a plan can be large
      updated and checked
                                                    Problem 3: Information
  Problem 4: Use of Plans                           "too little information"
The resulting plan is only useful if:
         a) the environment does not change
         b) the representation was accurate enough
         c) the robot's effectors are accurate enough to perfectly execute each step of
            the plan in order to make the next step possible
                      Reactive Systems
A purely reactive behavior-based method may be
represented by a horizontal (sequential) decomposition,
while the deliberative Sense-Plan-Act paradigm has
vertical (concurrent) decomposition.
                       Hybrid Systems
 Combine the two extremes
      reactive system on the bottom
      deliberative system on the top
      connected by some intermediate layer
 Layers must operate concurrently.
 Different representations and time-scales between the
A modern hybrid system typically consists of three components:
         a reactive layer
         a planner
         a layer that puts the two together.
=> Hybrid architectures are often called three-layer architectures.
            Behavior-Based Systems

 An alternative to hybrid systems
 Have the same capabilities
    the ability to act reactively
    the ability to act deliberately
 There is no intermediate layer
 A unified, consistent representation is used in the
  whole system => concurrent behaviors
 That resolves issues of time-scale.
• Regarding the architecture of robotic systems two key issues
  distinguishing architectures, as had to do with
    – time-scale (reactive) and
    – looking ahead (deliberative).

 • A third key issue we need to consider is modularity, i.e., the
 way in which the architecture decomposes into components.
     Task decomposition
          SPA architecture uses a functional (hierarchical ) decomposition

          Reactive architecture uses a task-oriented decomposition
          RCS uses a functional with different resolution (scale) in time and
         space at each level of the hierarchy.

• Behavior-based systems (BBS) use behaviors as the underlying
  module of the system, i.e., they use a behavioral decomposition.
                         Control Agents
A control agent or a control arbitration process is designed to
observe and respond to the (unstructured) environment in order
to fulfill a goal autonomously.
In doing so, it acts on behalf of a superior agent or human
The agent makes decisions in order to resolve conflicts between contradicting
 • An agent can be considered as a control subassembly.
 • Designing local behaviors for each agent that result in the
   desired global behavior (the control model) is a VERY
   hard problem.
     – Need for an architectural element to care about the Global Goal.
 • Let’s combine RCS with BBS
     – Behaviors implemented using control agents
                        RCS & BBS

• RCS is a type of hybrid system
   – Assumes that alternative plans, as response to contingent events
     can be pre-planned in relevant numbers.
• For some tasks, in special at lower layers with high control
  bandwidth, a behavioral decomposition will better suite.
   – Tasks may be complex only in the eye of an intelligent observer
• Behavior Generator Modules of each node in RCS already
  contain (execution) agents.
• A Behavioral Agent will suppress the preplanned
  execution and will be monitored directly by the superior
  agent in the hierarchy.
• Expected drawback: inability to precisely predict the
  output of the assemblage of behavioral agents.

To top