Robotic Control With Situation Aware Mobile Computing and Distributed Robot Agents

1
Robotic ControlWith Situation Aware Mobile
Computing and Distributed Robot Agents

• Brent Dingle
• Marco A. Morales
• Texas AM University, Spring 2002

2
Outline

• Definition of the problem
• A robot agent

3
Problem

• A robot is an Intelligent Connection of
  Perception to Action (Jones, Flynn 1993)
• The major problem in Robotic Motion Planning is
  stated thusTo plan an obstacle free path for a
  robot from a specified initial configuration
  (position, orientation, etc) to a specified final
  configuration.
• Multiple sensors and actuators
• Two main approaches
• Sequential based
• Behavior based

4
Control of Robots

• Two main approaches
• Sequential based
• Behavior based
• Little integration with the environment
• It is important to create autonomous robots, but
  technology could help to build robots competent
  enough for specific tasks.

5
Sequential Based Approach

• Sensors gather data
• Data are translated into a intermediate language
• A model of the world is built
• Motion planning is performed
• Motion commands are translated into low level
  orders for actuators

6
Behavior based approach

• Modules generate behaviors
• Each has perception and planning
• Each receives input and give commands
• A mediator scheme assigns control to modules
• Basic behaviors lead to complex behaviors
• No central model of the world
• No central control

7
Smaller Problems

• Currently robot control is done for SPECIFIC
  environments.
• Sequential approach
• It takes a great amount of time to find the
  solution to complex environments.
• The environment is often assumed static.
• Behavior based approach
• There is little knowledge to share about the
  environment
• The solution usually applies to a SPECIFIC robot.
• The solutions only deal with one robot.
• If the path doesnt work the robot gets stuck
  (until a human helps it out).

8
Small to One Situation Aware Mobile Computing
(SAMC)

• We would propose that these small problems can
  all be solved through the usage of techniques
  employed in Situation Aware Mobile Computing.
• SAMC obviously is related to Robotic Motion
  Planning.
• So we are going to assume some things
  (reasonably) so that we may incorporate the
  advantages of Situation Aware Mobile Computing
  into Robotic Motion Planning.

9
Assumptions

• Rooms exist with transensors devices that can
  send and receive wireless communications from
  mobile devices and relay them to a computer for
  processing.
• Robots are mobile devices.
• are equipped with a minimal set of functions
  defining how to move themselves about.
• are capable of translating a general command
  set into their hardware specific command set.
• Computers exist which are aware of
• Robots and have access to extra information on
  various types of robots and robot IDs.
• Various room states. Where the state of the
  room is derived from information relayed by
  transensors.
• General solutions to moving around the room,
  items in the room, and actions that may be taken
  on items in the room (and how to do so).
• The general command set and can relay directions
  for motion and action using this command set.

10
Picture (Proposed Idea)

• The solution usually applies to a SPECIFIC robot.
• No longer a problem as the solutions (paths and
  actions) are stored in the room in a generic
  language.
• Each robot becomes responsible for translating
  the general solution into specific commands.

11
Problems become solutions.

• The solutions only deal with one robot.
• Obviously since each robot is operating
  autonomously there is no loss of processing power
  to implement the solutions.
• And the problem is mostly pre-solved by the
  rooms distant computer.
• All the computer need do is send each robot a
  path in such a fashion as it will not collide or
  interfere with the path of another.
• The environment is often assumed static.
• As the transensors can track objects in the room
  the computer will know the location of all
  objects in the room even those moving.
• So if necessary small (and quick) adjustments can
  be sent by the rooms computer to correct for the
  dynamic environment (e.g. telling a given robot
  to delay 10 seconds so another may pass OR
  sending another robot to assist).

12
Problems become solutions.

• If the path doesnt work the robot gets stuck
  (until a human helps it out).
• Also no longer a problem.
• If a solution for whatever reason fails, the
  robot may send a request for another solution or
  request aid in performing its current task.

13
Extra Beauty

• As the rooms computer may be in contact with
  multiple robots at the same time (and objects in
  the room).
• It may direct robot B to help robot A, or to
  synchronize them together to achieve a task they
  cannot perform alone.
• This may involve sending robot A to find robot B
  known to be elsewhere in the building.
• Further the rooms computer could control access
  to the rooms or direct things such things as
  lifts or non-mobile robots (robot arms) to assist
  in accomplishing tasks.
• Robots can help each other to accomplish a task
  by sharing information only accessible to one of
  them at a time.

14
Robot Architecture
15
Components of the Robot Agent

• Planner
• Finds a path between two points or reports no
  such a path
• Navigator
• Creates a list of high level commands for the
  robot
• Pilot
• Gives low level commands to the controller and
  its aware of the sensors
• Controller
• Controls actuators in closed chains

16
Problems

• Environment modeling
• To model the environment each robot either is
  able to
• Identify main features by itself, or
• Uses a set of preloaded features.
• It seems reasonable to make robots use both.

17
More Problems

• Distribution of tasks
• Environments are far too complex for a robot to
  handle efficiently in detail
• A robot shouldnt care for parts of the
  environment that are far away, unless it really
  needs them.
• Nearby robots can help by providing info on the
  environment.
• Rooms, buildings, sites can help by having
  planning abilities.

18
Solution

• Build a distributed Motion Planning Agent
• Use nearby robots as sources of info for the
  environment
• Use precompiled info about the environment
• The local planner gathers info from all the
  robots in it and makes plans for them while they
  are nearby
• The robots can ask a local planner for a plan to
  follow
• The robots have the navigator, pilot and
  controller.
• A global planner coordinates the missions of the
  robots.

19
Distributed Planner

• Global planner
• Defines general goals based on main tasks
• Go to room 124, go to room 302
• Local planner
• Activated when the robot arrives to the area
  known to a given local planner
• Coordinates robot information with room
  information
• Stores a local map of the room and info gathered
  by all robots in the room
• Gives a plan to each robot in its influence
  area.

20
Robot agent

• Basic planner
• Takes control when no info is given by a room
  planner
• Navigator
• Pilot
• Controller

21
Advantage summary

• We have turned problems into solutions.
• We have introduced generality into the solutions.
• We have added the extra functionality of
  coordinating robots.
• We have decreased the functionality requirements
  of the robots (and likely the cost)
• We have increased the potential functionality of
  a given robot (no longer constrained to just a
  specific one or two tasks).

22
The newness

• Almost no robotic motion planning algorithms
  consider the possibility of another computer
  assisting in the path.
• Much of the research currently is on finding
  paths not accomplishing tasks.
• Coordinating actions through wireless
  communications would be an obvious direction to
  go in research.
• Building a generic command language to describe
  robotic motion needs to be done. (Aside Building
  a generic representation of rooms also needs to
  be done.)

23
Potential Future

• Incorporating humans into these concepts would be
  nice.
• This is already being done in medical operations.
• Could it be possible to design automatic
  assistants that coordinate in real time with the
  surgeon, via preplanned paths and a similar
  network.
• Could equipment auto relocate itself as needed
  from one surgery room to another based on calls
  from room computers and scheduled operations?

24
End of Talk

• Questions?
• Contact Brent Dingle
  dingle_at_cs.tamu.eduor Marco Morales
  marcom_at_cs.tamu.edu

25
Misc. thoughts

• Knowing the needs of the future is planning. )
• Preplanning paths is anticipating needs based on
  situation.
• Representing and recognizing the room state is
  not as important as representing and recognizing
  what just changed in the room.