The first phase

1
A Model-Driven Approach for Generating Embedded
Robot Navigation Control Software
Bina Shah (Iambina at- uab.edu) Rachael
Dennison (Raeanne at- aol.com) http//www.gray-a
rea.org/Research/CREW/ Advisor Dr. Jeff Gray
(gray at- cis.uab.edu) (www.gray-area.org)
An undergraduate research project, with support
from the NSF SIPHER program (Vanderbilt) and the
CRA Research Experience for Women (CRA-CREW)
2
Project Overview

• Goal
• Use of advanced modeling techniques to improve
  the development of embedded systems
• Synthesize robot control software from high-level
  models that depict configuration of a hostile
  environment containing robots, landmines, and
  lost babies

3
Motivating Problem and Solution Outline

• Motivating Problem
• Hard-coded control software for real-time
  embedded robotics control systems requires manual
  adaptation for each new configuration
• Solution Approach
• Use a meta-configurable modeling tool
• Create a meta-model that represents the hostile
  domain
• Construct a code generator that translates model
  information into robot control software
• Code generator has deep knowledge of domain and
  robot planning

4
Background

• The Lego Mindstorms Robotics Invention System
  (RIS)
• The RCX
• Programmable micro-controller embedded in a Lego
  brick
• IR Communication
• Software sends code to RCX using the IR tower
  connected to the serial or USB port
• RIS kit also includes
• Lego bricks, gears, sensors, motors, wheels

5
Model-Based Generators
Targets
Models stored as directed, attributed graphs
executable models
Synchronous Dataflow
Petri Net
Generators traverse/transform
analyzable model
6
Metamodeling and Modeling
MetaModel
Model

1. Abstract Concrete Syntax
2. Static Semantics
3. Visualization

OCL Constraints
7
Explanation of Hostile Grid Meta-Model

• Objects Baby, Landmine, Robot
• Attributes X-Coordinate, Y-Coordinate
• ConstraintsMinimum X-Coordinate, Minimum
  Y-Coordinate, Unique Name, Valid Name, Maximum
  Number of Robots, Unique set of X and Y
  coordinates

8
The Constraints Aspect
The Constraints MaxRobots lt 2 Xmin gt 0 Ymin
gt 0 UniqueName ValidName UniqueXYCoordinate
9
Example constraint

• UniqueXYCoordinate
• Pseudo code
• Return the number of babies, landmines, and
  robots with given X and Y coordinates.
• If number gt 1, the X and Y coordinate pair is not
  unique.

10

• Sample OCL Constraint
• let count project.allRobots(self.XCoordinate,
  self.YCoordinate) project.allBabies
  (self.XCoordinate, self.YCoordinate)
  project.allLandmines(self.XCoordinate,
  self.YCoordinate) in
• if (count lt 1) then
• true
• else
• false
• endif

11
Example Instance Model
12
Calculating Angle Rotations

• Rotation sensor reads a value of 16 when the
  wheel has rotated 360 degrees.
• Calculate the angle at which the robot needs to
  turn to point to so that it will travel to the
  baby in a straight line. double angle
  atan2((finalY-RobotY), (finalX-RobotX))
• Convert into degrees.
• angle angle (180/pi)

Rotation Sensors
13
A Model Interpreter for Generating Robot Control

• The Interpreter is written in C and hooks into
  modeling environment as a GME plug-in
• It will generate Java code
• The Java code will generate instructions for the
  robot to navigate to babies while moving and
  avoiding landmines

//Get the hostileGrid model const
CBuilderAtomList allRobots hostileDiagram-gtGetA
toms("Robot") pos1 allRobots-gtGetHeadPosition(
) CBuilderAtom Robot allRobots-gtGetNext(pos1)
//obtain the robot's (X,Y) coordinates--gt
(RobotX, RobotY) int RobotX, RobotY Robot-gtGetA
ttribute("XCoordinate", RobotX) Robot-gtGetAttrib
ute("YCoordinate", RobotY) //Get the hostileGrid
model const CBuilderAtomList allRobots
hostileDiagram-gtGetAtoms("Robot") pos1
allRobots-gtGetHeadPosition() CBuilderAtom
Robot allRobots-gtGetNext(pos1) //obtain
the robot's (X,Y) coordinates--gt (RobotX,
RobotY) int RobotX, RobotY Robot-gtGetAttribute(
"XCoordinate", RobotX) Robot-gtGetAttribute("YCoo
rdinate", RobotY) //Get the hostileGrid
model const CBuilderAtomList allRobots
hostileDiagram-gtGetAtoms("Robot") pos1
allRobots-gtGetHeadPosition() CBuilderAtom
Robot allRobots-gtGetNext(pos1) //obtain
the robot's (X,Y) coordinates--gt (RobotX,
RobotY) int RobotX, RobotY Robot-gtGetAttribute(
"XCoordinate", RobotX) Robot-gtGetAttribute("YCoo
rdinate", RobotY) //Get the hostileGrid
model const CBuilderAtomList allRobots
hostileDiagram-gtGetAtoms("Robot") pos1
allRobots-gtGetHeadPosition() CBuilderAtom
Robot allRobots-gtGetNext(pos1) //obtain
the robot's (X,Y) coordinates--gt (RobotX,
RobotY) int RobotX, RobotY Robot-gtGetAttribut
e("YCoordinate", RobotY)
ModelInterpreter(C)
Generated Control Code
14
Snippet of Interpreter Code
Obtaining robot coordinates from model
//Get the hostileGrid model const
CBuilderAtomList allRobotshostileDiagram-gtGetAto
ms("Robot") pos1 allRobots-gtGetHeadPosition()
CBuilderAtom Robot allRobots-gtGetNext(pos1)
//obtain the robot's (X,Y) coordinates--gt
(RobotX, RobotY) int RobotX, RobotY Robot-gtGetA
ttribute("XCoordinate", RobotX) Robot-gtGetAttrib
ute("YCoordinate", RobotY)
15
Limitations of the robot

• Sensitive to light
• Inaccurate angle calculation
• One byte payloads
• Works only in line of sight

16
Summary

• Hard-coding models to adapt to new
  configuration can be automated using
  meta-configurable modeling tool.
• A meta-model is created that allows to create
  a hostile
• environment including lost babies, landmines,
  and robots.
• The code generator, interpreter, is written in
  C, which translates the model information into
  robot control software.

17
Questions?

• http//www.gray-area.org/Research/CREW/