Transcript presentation source

Study of noise readings in
infrared sensors and their effect
in the Khepera Miniature Robot’s
performance
Saúl J. Vega
Daisy A. Ortiz
Advisor: Raúl E. Torres, Ph.D., P.E.
Background
• Robots
–
–
–
–
Management of hazardous waste
Moving of heavy equipment
Ocean and space exploration
Fire extinguishing
• Artificial Intelligence
– Knowledge-based
– Behavior-based
Background (cont.)
• Behavior-based Artificial Intelligence
– Subsumption Architecture (SA)
• Build behaviors from smaller sub-behaviors
• SA rely heavily on sensory input
– Noise cause disturbance in robot
operation
Problem Statement
• Avoid negative effect of fluorescent
lamps on infrared sensory readings
Oh! No!
Objectives
• Determine the effect of noisy readings
on robot performance
• Determine the effect of filtered sensory
on robot performance
Methodology
• Review of literature
• Simulation study
• Hardware implementation
– Real Khepera used in testing
• Filters design
• Testing-platform development
– Braitenberg vehicle algorithm
• Comparison of results
Our Star:
Mr. Khepera
Anatomy of the Khepera
Microprocessor
IR-Sensors
Wheels & DC-Motors
Insights of Our Hero
Microprocessor
IR-Sensors
Wheels & DC-Motors
Simplified Braitenberg Algorithm
Obstacle on
Left side?
No
Yes
Obstacle on
Right side?
No
Move Forward
Yes
Turn Right
Turn Left
No
Obstacle on
Back?
Yes
End
n
1002
353
331
925
265
243
221
199
177
155
133
111
89
67
45
23
1
309
1200
1000
800
600
400
200
0
848
No filter in Light
287
n
771
694
617
540
463
386
309
232
155
78
1
IR-values
IR-values
No filter in Darkness
1200
1000
800
600
400
200
0
n
221
243
265
221
243
265
155
133
111
89
67
45
23
1
199
1200
1000
800
600
400
200
0
199
Averaging in Light
177
n
177
155
133
111
89
67
45
23
1
IR values
IR values
IIR filter in Light
1200
1000
800
600
400
200
0
Results
• On darkness
– Slow filter response when approaching obstacle
– Even slower when moving away from obstacle
• On light
– Acceptable filter response time when
approaching obstacle
– Acceptable filter response time when moving
away from obstacle
– Noisy readings greatly reduced
Results (cont.)
• Satisfactory performance of Braitenberg
algorithm without filtered readings on
darkness
• Problems using filters with Braitenberg
algorithm
– Robot slow to react to filtered sensory readings
Conclusions
• Fluorescent light noise cause serious effects
on Khepera’s performance
• Digital filters proved to be useful in
reducing noise in sensory readings
• Filters performance are greatly affected by
levels of ambient light
Future Works
• Braitenberg algorithm modified to allows
detection of ambient light
– Activate filters on high levels of ambient light
– Disable filters on low-light conditions
• Develop user-friendly program for testing
algorithms and filters
Questions?