2005CON Control Communication Using ROBOLAB

Uploaded from authorPOINTLite
Views:
 
Category: Entertainment
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Advanced Robot Control: 

Advanced Robot Control Programming for Robustness with RoboLab

Positioning: 

Positioning Absolute Uses features or ‘landmarks’ of the course Relative Robot keeps track of its moves Relies on Odometry

Positioning Problems: 

Positioning Problems Absolute May have difficult time finding small landmarks Some landmarks & robots are easily damaged Relative Error accumulates with every move If too many errors, robot maybe too far off course to find landmark later

Common Sources of Error: 

Common Sources of Error Rotation Sensor Resolution Gear Backlash Program Execution Speed Wheel Spin/Skidding Inside Turn Wheel

Rotation Sensor Resolution: 

Rotation Sensor Resolution Robot only knows position with plus or minus one count (at best) Gear backlash increases error beyond one count Use finer resolution to reduce error (Minimize Distance per Count) Rotation sensor should be at same speed as motor (or up to 1-1/2 times higher)

Program Execution Speed: 

Program Execution Speed Rotation sensor not read continuously RCX may not ‘see’ a target RCX will not react instantly

Wheel Spin at Startup: 

Wheel Spin at Startup Caused by sudden application of motor torque, not enough weight on drive wheels Wheels and rotation sensor turn before robot starts Skip or changes direction due to “jump” from start

Skidding: 

Skidding Caused by rapid application of motor braking and not enough weight on drive wheels Robot told to stop but continues to move Rotation sensor doesn’t ‘see’ move Sends robot off position, affecting next move by robot

Turns: 

Turns Errors are magnified in turns Any slight direction error can cause larger side-to-side error Braking of inside wheel Any movement of the inside wheel lessens the overall turn; true angle is shorter than with a locked wheel Turns made with two counter-rotating wheels doubles rotation sensor resolution errors Additional errors if wheels don’t turn at same speeds

Non-Programming Solutions: 

Non-Programming Solutions Set a reasonable speed-Try gearing robot for 10 to 15 inches per second Allows one wheel to be ‘locked’ in turn Gear rotation sensor for 1/8” of travel per count or less Measures position as precisely as practical Minimize backlash by avoiding multi-stage gearing Avoid loose gear meshes Keep weight on driving wheels Gain traction Minimize slipping and skidding Weight shifts with accel/decel Match motors – use two motors with same output speeds Use motor test jig

Motor Test Jig: 

Motor Test Jig Build a motor test jig using: Load motor with worm geartrain Test and record motor data Run for turn, record counts Forward and reverse Different power levels Picture of Motor Test Jig

Programming Solutions: 

Programming Solutions ‘Creeping’ ‘Precise’ Forward/Reverse/Turns ‘Square Up’ to Lines Line Following using ‘shades of gray’ Experimentation

‘Creeping’: 

‘Creeping’ Moves Robot Slowly by providing a series of taps to the robot Overcome static friction Provides braking and speed control Offers these Advantages Go slowly to minimize wheel slippage Minimize distance error due to polling error Better chance of sensing narrow lines Bump up against landmark with much less force

Why Not Use Low Power Levels?: 

Why Not Use Low Power Levels? Often don’t provide enough power to overcome static friction Robot still rolls easily enough that speed is still too high

How to Creep: 

How to Creep Create a loop to wait for rotation (or time, light level or button press) Start motors at medium power level Wait for a very small time (1/100 sec) Stop the motors Wait for a very small time (1/100 sec) End loop

Creep Example: 

Creep Example

Precise Turns/Forward/Reverse: 

Precise Turns/Forward/Reverse Power applied gradually Reduce power before target Creep forward/backward until reach target

‘Precise’ Startup : 

‘Precise’ Startup Uses subroutine (to save memory) Position target passed from main task via container Sets up intial target Try using 10 to 20 counts short of actual target Loops until initial rotation target Branches to different power levels based on timer to provide smoother acceleration Avoids wheel slip at startup

At Initial Target: 

At Initial Target Coast or Creep If coasting, coast until time Could possibly coast past target Creeping applies pulsed braking No skidding Self correcting using closed loop positioning Moves forward or reverse to final target count Too far – creeps in reverse Too short – creeps forward

Routine Details: 

Routine Details One subroutine can be used for left turns and forward Container 7 is set to 0 or 1 to choose left or forward Reverse or right turns are done similarly Stored as subroutines to save memory Target counts are passed using blue container Set container for forward/reverse or left/right

Square Up: 

Square Up Line up robot to edge of line Uses two Light Sensors Moves robot so each sensor seeks dark/light edge Know exact spot when parked Accuracy in direction Accuracy in position (1 axis)

Square Up ‘Setup’: 

Square Up ‘Setup’ Square up to dark line Each sensor is different Needs to be set before running Separate sub-vi that calibrates light levels Grabs light values Calculates and stores threshold values

How it works: 

How it works For each sensor: If sensor sees: Light: Creep one pulse forward, Reset Container to 0 Dark: Add 1 to container Do until container is set to 2 which means both sensors made it to the dark line Robot “waddles” to the line Repeat process with motors set for reverse and looking for light instead of dark Repeat loop two times to assure exact placement

Line Following: 

Line Following Follow line edge using light sensor Reads average value of light within a circle Seeking halfway between light and dark Based upon Light level sensed Motor ‘Behavior’ will set motors to creep to steer robot toward line edge Can be separated into the 7 zones (‘shades of grey’) Can go straight or turn depending on value Go faster and straighter near middle zones Go slower and turn sharper in zones away from middle

Program Example: 

Program Example Create an outer loop rotation sensor target Create a decision tree within the loop Made with container forks for branching for different response to each light level range Use Creeping within each branch Each of the 7 conditions can be setup and tested individually

Experimentation is Key: 

Experimentation is Key Alter creep speed and turn radius Watch robot to see how it behaves Adjusting height of light sensor Changes size of circle being read Changes sensitivity Adjust location of light sensor Change weight distribution

Memory Management: 

Memory Management Use Subroutines(‘Subs’) for routines called repeatedly Pass parameters to ‘Subs’ using Containers Use Containers as flags (for program forks) to get multiple functions per Sub. Use utility programs to show memory usage and clear out slots. Get to know memory usage of program elements

Show Memory vi example: 

Show Memory vi example

Erase Slot vi example: 

Erase Slot vi example

One Final Thought:: 

“The lesson is in the struggle and not in the victory” One Final Thought: