Evaluation of Sensors for a Smart Wheelchair
Edmund F. LoPresti & Richard C. Simpson, University of Pittsburgh, PA
David Miller, Kiss Institute for Practical Robotics, Norman, OK
Illah Nourbakhsh, Carnegie Mellon University, Pittsburgh, PA
ABSTRACT
Slide text:
- Seven sensors were evaluated to determine their ability to detect obstacles at different distances and angles.
- These included six sonar sensor models and two infrared sensor models.
- Each sensor was evaluated for eight materials and four obstacle widths.
- Results of this analysis are being used to equip a wheelchair that provides navigation assistance.
SENSOR MODELS
Slide text:
- Polaroid 600 sonar transducer (Polaroid, Cambridge, MA);
- Polaroid 9000 wide-beam sonar transducer;
- Sharp GP2D02 digital output infrared sensor (Sharp Electronics, Mahwah, NJ);
- Sharp GP2D12 analog output infrared sensor;Massa E-152/40 sonar sensor (Massa Electronics, Hingham, MA);
- Sonaswitch Mini-A sonar sensor (EDP Company, Livonia, MI).Devantech SRF04 (Robot Electronics, Norfolk, UK)
PERFORMANCE MEASURES
Slide text:
OBSTACLES
Slide text:
Detection distance and detection angle measured for large, flat obstacles, for 8 materials:
- Bare Drywall
- Black Drywall
- White Drywall
- Wood
- Brick
- Marble
- Carpeting
- Brick
Detection angle was measured for thin, cardboard obstacles for 8 obstacles widths:
- 1.25cm
- 2.5cm
- 5.0cm
- 10.0cm
MINIMUM DETECTION DISTANCE (Median Across Materials)
Graphic description:
The graph for minimum detection distance shows the shortest distance for which each sensor could reliably detect an obstacle. For each sensor model, the median is shown across all obstacle materials. The values are:Polaroid 600: 8.2 cm; Polaroid 9000: 25.6 cm; Sharp GP2D02: 3.0 cm; Sharp GP2D12: 0.8 cm; Massa: 34.6 cm; Sonaswitch: 21.8 cm; Devantech: 0.8 cm.
The graph shows that the minimum distance was quite small for the Devantech sonar and both Sharp infrared sensors, slightly farther away for the Polaroid 600, and noticeably farther for the Polaroid 9000, Massa, and Sonaswitch.
MAXIMUM DETECTION DISTANCE(Median Across Material)
Graphic description:
The graph for maximum detection distance shows the largest distance for which each sensor could reliably detect an obstacle. For each sensor model, the median is shown across all obstacle materials. The values are:Polaroid 600: 348.9 cm; Polaroid 9000: 405.3 cm; Sharp GP2D02: 63.2 cm; Sharp GP2D12: 75.6; Massa: 248.3 cm; Sonaswitch: 238.2 cm; Devantech: 183.8 cm.
The graph shows that the two Polaroid sensors were able to detect obstacles at the greatest distance, followed by the Massa, Sonaswitch, and Devantech, and finally the two Sharp infrared sensors.
RESULTS
Slide text:
- Infrared sensors able to detect closer obstacles and detect large obstacles at greater angles.
- Ultrasound sensors able to detect obstacles at a larger distance.
- Polaroid 9000 unable to reliably detect smallest obstacles (1.25 cm width).
- Material effects most noticeable for infrared sensors, which had difficulty detecting black drywall, glass, and marble (significantly larger minimum and smaller maximum detection distances, p<0.05, for GP2D12).
DETECTION ANGLE FOR LARGE OBSTACLES (Median Across Materials)
Graphic description:
The first graph shows Detection Angle, the angle at which each sensor could detect a large obstacle (between 40 and 150 cm wide). For each sensor model, the median is shown across all obstacle materials.
The values are:Polaroid 600: 56.9 degrees; Polaroid 9000: 66.2 degrees; Sharp GP2D02: 141.7 degrees; Sharp GP2D12: 141.2 degrees; Massa: 43.6 degrees; Sonaswitch: 68.0 cm; Devantech: 79.4 degrees. The graph indicates that the two Sharp infrared sensors could detect obstacles at the greatest angles.
MAXIMUM DISTANCE ACROSS MATERIALS, SHARP GP2D12
Graphic description:
The second graph shows the effect of obstacle material properties on detection distance for the Sharp GP2D12 infrared sensor. The detection distances were bare drywall: 98.9 cm; white drywall: 115.8 cm; black drywall: 30.9 cm; wood: 100.2 cm, glass: 28.2 cm; carpet: 89.8 cm; marble: 39.2 cm; brick: 61.4 cm. The graph indicates that infrared detection distance was lowest for black drywall, black marble and glass.
CONCLUSION
- Using Sonaswitch ultrasound, Sharp GP2D12 for Smart Wheelchair development.
- Ultrasound & infrared sensors will be used together for redundancy.
- Relationship between sensor output and obstacle distance will be used to estimate distance to obstacles; sample relationships for two sensors are shown below.
- One limitation of this evaluation is that all obstacles were stationary relative to the sensors; a smart wheelchair will be moving relative to obstacles.
SONASWITCH VOLTAGE-DISTANCE RELATIONSHIP
Graphic description:
The third graph shows Sonaswitch Voltage-Distance Relationship. The Sonaswitch sensor voltage response is near zero until eight inches, after which it rises in an almost linear manner until an object is 38 inches away, at which point the voltage remains near five volts.
SHARP GP2D12 VOLTAGE-DISTANCE RELATIONSHIP
Graphic description:
The fourth graph shows Sharp GP2D12 Voltage-Distance Relationship. The infrared sensor reading rises steeply from zero volts at zero inches to almost three volts at three inches, then falls asymptotically until it reaches zero volts near 42 inches.
The End
Return to Slide Series
Updated: May 6, 2003
|
