Development of Industrial Picking System

Decrease of the population Reality on Robots

Estimate of the population on Kitakyushu City

Agriculture

Food

Garbage disposal

Medical

The working population will decrease.

Very complicate

Not Smart

Strong• Exclusive controller

• Teaching is needed

• Almost everyone can not use

• Only professional engineers

can deal with.

It’s difficult to operate industrial robots.

We would like to solve with Industrial Robots. Industrial robots should be smarter.

3D sensors

Measurement of

objects,

robots,

and surroundings info

Autonomous

Industrial Robot

Humanoid Robot

Human Interface

to control industrial robots

Measurement of

objects, robots,

and surroundings info

3D sensors

Human Interface

To control industrial robots

Calculate motion plan

And picking plan

autonomously

Autonomous

Industrial RobotHumanoid Robot

１．Development of the UI

２．Robotic Intelligent Space

• Humanoid Robot

• Graphical UI

３．Smart industrial robot

• Multiple 3D sensors measures surroundings

• No occlusion

Autonomous calculation of the motion plan

Generated motion

3D sensors

Robotic Intelligent SpaceGraphical UI

Easy operation of industrial robots.

Ryodo TANAKA, Ryo KABUTAN, Shinji OOMORI, Masaru MORITA, Takeshi NISHIDA

Robotic Intelligent Space + Smart Industrial Robot

① ②

Set the static environment Calculate the motionput objects detected objects

Bounding box informationare output.

2

Output

Projecting 3D points on

2D points

1 Measurement of the

point cloud data

Searching minimum

bounding rectangle

3 Finding a rectangle.

• We’ve used motion planning library.

• Motion planning algorithm

is RRT-Connect.

Image of RRT-Connect

Fast Calculation

Practical test & example

Result Count Rate[%]

Success 100 83.3

Motion planning

failure 13 10.8

D-Hand error 4 3.33

Gripping failure 2 1.67

Object detection

error1 0.833

43

44

2 3

Number of failures in each areas

Repeating the experiment to execute

pick and place tasks 120 times.

The success rate of motion planning

should be increased.

Proposed method + IBM Watson Recognition

Check the result of the learning

Watson API

box API

Cloud

Factory

Human

Intelligence

①②

③

③④ ④

④

⑤

⑤

IBM Bluemix Hackathon

We won the 2nd/80 price !!

= A lot of people

NEEDS our system.

Bounding box information is output.

Actual value Estimate value

Over 65 age 15 ~ 64 age Under 14 age Rate of aging

Actual value Estimate value

Over 65 age 15 ~ 64 age Under 14 age Rate of aging