Robot Swarms, Lego Bricks, Ant Farms, and Star Wars: How ... Conferencia-Magistral... · Decision...

50
Swarm Logistics Why a Thousand Robots are Better Than One James McLurkin

Transcript of Robot Swarms, Lego Bricks, Ant Farms, and Star Wars: How ... Conferencia-Magistral... · Decision...

Swarm Logistics

Why a Thousand Robots are Better Than One

James McLurkin

Part I: The Robot Revolution has Already Begun (And it is being televised)

iRobot Roomba

picture courtesy iRobot

Honda ASIMO

picture courtesy Honda

The iRobot Packbot

pictures courtesy iRobot

The “Three D’s” of Robotics:

Dangerous

Dirty

Dull

Distributed

Amazon.com: Warehouse Distribution

Large populations of robots can fundamentally transform the problem

Why Robots?

The reasons to use robots are well known:

•Cost Savings

•Precision

•Efficiency

Multi-Robot systems provide even more advantages

•Parallel Operations

•Geographic Distribution

•Redundant, Fault-Tolerant operation

Multi-Robot Systems can Transform Logistics Problems

(What does that mean, anyway?)

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Logistics in Challenging Environments

What if we sent

20 robots to look for hot spots in forest fires?

What if we sent

200 robots to look for survivors after an

earthquake?

What if we sent

2,000 robots to Explore Mars?

Part II: Distributed Algorithms 101

But use a Distributed Algorithm!

Algorithm: n: a procedure for solving a mathematical problem.

Distributed: adj. to divide among several or many.

Distributed Algorithm: n. A program that runs on multiple computers that interact to form a group result

Computation Model

Computation Model

Local Communications

Each robot can communicate with robots within range

r=

Global Network

r=

Broadcast Tree Construction

source

Intanagonwiwat, Govindan, Estrin. “Directed Diffusion: A Scalable and Robust Communication Paradigm for

Sensor Networks”, 2000

Part III: Multi-Robot Systems and Logistics Problems

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Building Search

GuideBot

ChargingBot

InteriorBot

BoundaryBot

1. Disperse throughout a building

2. Find an item of interest

3. Lead the user to the item

Recovery with “Bread Crumbs”

“Leaf” robots:

Goal

Internal robots:

We want the maximum

number of leaf robots,

i.e. the maximum leaf

spanning tree

Recovery with “Bread Crumbs”

2-DMLST Recovered 97.2% of 17 robots in 9 trials • Hardware issues/physical interference.

• One case of odd communications failure

No breadcrumbs 56 seconds, 10 lost robots

Breadcrumbs 6min 32 sec, no lost robots

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

If you have enough robots, you can transform the problem of measuring the environment…

Boundary Detection

…to one of measuring the

robots

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Physical Bubble Sort

Goal: Sort the robots by their robot ID

5

2

9

Physical Bubble Sort

Goal: Sort the robots by their robot ID

2

9

3

5

7

Physical Bubble Sort

Goal: Sort the robots by their robot ID

2

9

3

5

7

Physical Bubble Sort

Goal: Sort the robots by their robot ID

2

9

3

5 7

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Multi-Robot Manipulation

Multi-Robot Manipulator Design

Multi-Robot Manipulation: Gripper

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Seeley, T.D., 1995. The Wisdom

of the Hive. Harvard University

Press: Cambridge, MA

Nectar Collection in Honeybees

Honeybee Computation

Honeybee workers share food all the time, computing a global average. This lets an individual worker know

when the hive is hungry by measuring when she is hungry.

Time step person 1 person 2 person 3 person 4 person 5 person 6 person 7 person 8

1 30.0 10.0 90.0 110.0 40.0 120.0 80.0 40.0

2 20.0 20.0 100.0 100.0 80.0 80.0 60.0 60.0

3 40.0 60.0 60.0 90.0 90.0 70.0 70.0 40.0

4 50.0 50.0 75.0 75.0 80.0 80.0 55.0 55.0

5 52.5 62.5 62.5 77.5 77.5 67.5 67.5 52.5

6 57.5 57.5 70.0 70.0 72.5 72.5 60.0 60.0

7 58.8 63.8 63.8 71.3 71.3 66.3 66.3 58.8

8 61.3 61.3 67.5 67.5 68.8 68.8 62.5 62.5

9 61.9 64.4 64.4 68.1 68.1 65.6 65.6 61.9

10 63.1 63.1 66.3 66.3 66.9 66.9 63.8 63.8

11 63.4 64.7 64.7 66.6 66.6 65.3 65.3 63.4

12 64.1 64.1 65.6 65.6 65.9 65.9 64.4 64.4

13 64.2 64.8 64.8 65.8 65.8 65.2 65.2 64.2

14 64.5 64.5 65.3 65.3 65.5 65.5 64.7 64.7

15 64.6 64.9 64.9 65.4 65.4 65.1 65.1 64.6

16 64.8 64.8 65.2 65.2 65.2 65.2 64.8 64.8

17 64.8 65.0 65.0 65.2 65.2 65.0 65.0 64.8

18 64.9 64.9 65.1 65.1 65.1 65.1 64.9 64.9

19 64.9 65.0 65.0 65.1 65.1 65.0 65.0 64.9

20 64.9 64.9 65.0 65.0 65.1 65.1 65.0 65.0

21 65.0 65.0 65.0 65.0 65.0 65.0 65.0 65.0

Simulation Results:

0

10

20

30

40

50

60

70

80

90

100

110

120

1 6 11 16 21 26 31 36

person 1

person 2

person 3

person 4

person 5

person 6

person 7

person 8

Partial Proof:

Multi-Robot Systems Transform Logistics Problems

Problem Multi-Robot

Transformative Property Transformed Problem

Exploration Population Dispersion Exploration

Mapping Local Network Geometry Boundary Detection

Organization Parallel Operation Sorting

Collaboration Scalable Response Ant-Like Manipulation

Decision Making Distributed Computation Honey Bee Economics

Part IV: Conclusion

Summary

1. Robots are here, and they’re useful

2. Multi-Robot Systems offer unique advantages for Logistics

3. Multi-Robot Systems require different techniques for control

4. Multi-Robot Systems can provide transformative solutions

end