(sensitivity to falling snow) 1 Sbastien Michaud, 2 JeanFranois - - PowerPoint PPT Presentation

sensitivity to falling snow
SMART_READER_LITE
LIVE PREVIEW

(sensitivity to falling snow) 1 Sbastien Michaud, 2 JeanFranois - - PowerPoint PPT Presentation

(sensitivity to falling snow) 1 Sbastien Michaud, 2 JeanFranois Lalonde, 1 Philippe Gigure 1 Computer Science and Software Engineering 2 Electrical and Computer Science Laval University, Quebec City, Canada 7 th Workshop on Planning,


slide-1
SLIDE 1

1Computer Science and Software Engineering 2Electrical and Computer Science

Laval University, Quebec City, Canada 7th Workshop on Planning, Perception and Navigation for Intelligent Vehicles IROS, 2015

1Sébastien Michaud, 2Jean‐François Lalonde, 1Philippe Giguère

(sensitivity to falling snow)

slide-2
SLIDE 2

Motivation and context

2/16

Atlanta, 2014 6 cm snow Rolling Dead scenario when snow is falling?

Walking Dead

slide-3
SLIDE 3

Outline

  • List of LiDARs + description
  • Physical test bench for data collection
  • Analysis of snowflake echoes in LiDARs:

– Temporal variation during snowstorms – Probabilistic sensor modeling as function of range (for Bayesian perception)

  • Discussion and conclusion

3/16

slide-4
SLIDE 4

Description of LiDARs

4/16

Sensor Maximum distance (m) Spot Area at 30 m (cm²) Spot Shape Number

  • f echos

SICK LMS200 28 165 Circle 1 SICK LMS151 50 22 Circle 2* Hokuyo UTM‐30LX‐EW 30 196 Ellipse 3 Velodyne HDL‐32E 70 51 Rectangle 1 Velodyne on loan from Canadian Space Agency UTM‐30LX‐EW was on loan from Hokuyo *was programmed to return last echo

slide-5
SLIDE 5

Physical setup : 4 sensors in a row

5/16

  • Provide some

shielding to avoid sensor clogging

  • Fast to deploy

(just open the window!)

  • Simultaneous

data collection

  • Could be left
  • pen for long

periods of time

slide-6
SLIDE 6

View of the “target”

  • Target away from pedestrians and trees

6/16

pedestrian walkway Approximate Lidar Impact location range between 14 m to 22 m

slide-7
SLIDE 7

Variety of snowfall episodes gathered

  • Total: 50 hours of recording (>300 Gb)
  • The ones in bold were used in the study
  • rosbags available upon demand

7

slide-8
SLIDE 8

Sensor data gathering settings

8/16

(away from walkways)

slide-9
SLIDE 9

Snowflake detection + temporal smoothing

  • Threshold for echo: 14 m
  • To simplify viewing,

used averaging windows (remove very short term dynamics)

9

Clusters of echoes for Hokuyo

  • thresh. for echoes
  • thresh. for echoes

Repetitions unlikely

slide-10
SLIDE 10

Snowstorm temporal evolution (1)

10/16

SICK LMS200 (1x) Hokuyo 1st echo (1x) Veolodyne (30x) Hokuyo last echo (30x)

Similar behavior Similar behavior

Note: (‐‐x) is scaling factor

LMS151 (200x)

slide-11
SLIDE 11

Snowstorm temporal evolution (2)

11/16

Numerous fluctuations during a storm Average over this event

slide-12
SLIDE 12

Snowstorm temporal evolution (3)

12/16

  • Snowstorms have different

temporal evolutions

  • Often have sharp peaks

Peak avg 0.5% Velodyne ~2cm/h snow

slide-13
SLIDE 13

Probabilistic model of echoes P(Echo|x)

  • Distributions normalized with area = 1

13

model

Overall %

  • f echoes

Shielding would be much less on a vehicle

slide-14
SLIDE 14

Physical interpretation

  • Large targets: return is ~
  • Snowflakes: return is ~

14/16

2

1 d

(most of the beam paints target)

Sensor Spot Area at 30 m (cm²) SICK LMS200 165 SICK LMS151 22 Hokuyo UTM‐30LX‐EW 196 Velodyne HDL‐32E 51

(fraction of beam intercepted is 1/d²)

4

1 d

slide-15
SLIDE 15

Conclusion

  • Snowstorms are highly dynamic
  • Most recent LiDARs have very low probability of

snow interference (<0.1%)

  • Multi‐echoes can be used to detect snowing

conditions

  • Distribution follows more‐or‐less a log‐normal as

function of range: easy to put in a simulator

  • Snowflakes are virtually undetectable beyond 10 m

(should be able to still use building walls for localization)

15/16

slide-16
SLIDE 16

Future work

  • % of missing echoes
  • Impact of sunlight
  • Noise on target
  • Intensity of echoes in snowflakes as function of

distance

  • Collect data in a more car‐like setup

16/16