s u m m i t - amazon web services...video streams, amazon rekognition, and amazon cloudwatch lex...

Post on 22-May-2020

8 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

S U MM I TB E R L I N

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Rise of Robotics:Innovating the Future with IoT & AI

Özkan Can, AWS - Senior Solutions Architect

@_ozzc

A W S S U M M I T B e r l i n

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Agenda

• Cloud Robotics

• Robot Development

• Robot Collaboration

• Reinforcement Learning

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Robotics trendsin 2018

Robotics is undergoing fundamental change in collaboration, autonomous

mobility, and increasing intelligence

Source: IDTechEx

• Logistics

• Construction

• Retail

• Hospitality

• Healthcare

Robots are being put to work every

day across many industries

• Agriculture

• Energy Management

• Oil and Gas

• Facilities Management

• Household chores

By 2023, it’s estimated that mobile autonomous robots will emerge as the standard for logistic and fulfillment processes

By 2030, 70% of all mobile material handling equipment will be autonomous

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Mobile roboticstrends 2018

We are at the beginning of an inflection point, where expected growth in the use of mobile robots will increase by almost tenfold over the next 2–3 years, but…

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Cloud Robotics

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Robotic development is difficult and time consuming

Requires machine learning expertise for intelligent functions

Many prototyping iterations

Days spent setting up and configuring

Months to building a realistic simulation environment

Duplicated efforts integrating an application management system

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Robotic development cycle

2) Develop robotics

application

1) Select robotics software

framework

4) Deploy and manage

application

3) Test and simulate

application

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS RoboMaker

A service that makes it easy for developers to develop, test, and deploy robotics applications, as well as build intelligent robotics functions using cloud services

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS RoboMaker service suite

Development Environment

Simulation Cloud Extensions for ROS Fleet Management

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Robot Operationg System (ROS) Primer

Robot Hardware

OS

ROS

Robot Application

ROS is an open-source,

meta-operating system for your robot.

https://wiki.ros.org

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

ROS Nodes

ROSMaster

ROSNodes

/topicsROS

NodesROS

NodesROS

Nodes

ROSNodes

ROSNodes

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

ROS Nodes

ROSMaster

ROSNodes

/topicsROS

NodesROS

NodesROS

Nodes

ROSNodes

ROSNodes

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Hello World Node Publisher

1 import rospy

2 from std_msgs.msg import String

3 pub = rospy.Publisher('chatter', String, queue_size=10)

4 rospy.init_node('talker', anonymous=True)

5 rate = rospy.Rate(10) # 10hz

6 while not rospy.is_shutdown():

7 hello_str = "hello world %s" % rospy.get_time()

8 pub.publish(hello_str)

9 rate.sleep()

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Hello World Node Subscriber

1 import rospy

2 from std_msgs.msg import String

3 def callback(data):

4 rospy.loginfo(rospy.get_caller_id() + "I heard %s", data.data)

5 rospy.init_node('listener', anonymous=True)

6 rospy.Subscriber("chatter", String, callback)

7 rospy.spin()

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS RoboMaker Investment in ROS

Actively guiding the development of the newest version of ROS (ROS2) via representation on the Technical Steering Committee (TSC) of OSRF

Investment to provide Long Term Support (LTS) for ROS2

• LTS maintains backwards compatibility

• LTS allows developers to release commercial and production-level products on ROS2

Investment to increase ROS2 packages

• Packages are reusable and provide unique robotics functions such as mapping and navigation

• AWS driving migration of 113 critical packages identified for ground mobility robots to ROS2 by the first ROS2 LTS release

Leading the security and logging design of ROS2 and contribution to the ROS2 core messaging system, Data Distribution Service (DDS)

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS RoboMaker Architecture

ROS/ROS2

DevelopmentEnvironment

SimulationFleet

ManagementCloud Extensions

for ROS

Bu

ild

To

ol

Clo

ud

De

ve

lop

me

nt

En

vir

on

me

nt

De

bu

g

ML

Mo

de

l T

rain

ing

Ite

rati

ve

S

imu

lati

on

Re

gre

ssio

n

Te

stin

g

Re

gis

tra

tio

n

De

plo

ym

en

t

Mo

nit

ori

ng

Ma

na

ge

me

nt

Le

x

Po

lly

Re

ko

gn

itio

n

Kin

esi

s V

ide

o S

tre

am

s

Clo

ud

Wa

tch

Me

tric

s

Clo

ud

Wa

tch

Lo

gs

Me

tric

an

d L

og

gin

g

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS RoboMaker

Cloud Extensions for ROS

Cloud extensions written as ROS packages automatically create connections and make API calls to AWS services, such as Amazon Lex, Amazon Polly, Amazon Kinesis Video Streams, Amazon Rekognition, and Amazon CloudWatch

LEXspeech

recognition

POLLY

speech

generation

KINESIS

VIDEO

STREAMS

video

streams

REKOGNITION

image and video

analysis

CLOUDWATCH

logging and

monitoring

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Amazon CloudWatch extension in operation

ROS Nodes

and topics

AWS Cloud

Node logic

Temporary

security

credential

Stored

credential

Robot

… or …

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Collaboration through IoT Services

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

To securely connect devices to the AWS cloud & other devices at scale

To fully integrate with other AWS services to reason

on top of the data (Analytics, Databases, AI, etc.)

To route, process, and act upon data from connected devices

AWS IoT Core is a managed service that lets connected devices easily

and securely interact with cloud applications and other devices.

To enable applications to interact with devices even

when they are offline

AWS IoT Core

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Edge Cloud

Law of Economics

Law of Physics

Law of the Land

AWS IoT Greengrass

AWS IoT Greengrass extends AWS IoT onto your devices, so that they can act

locally on the data they generate, while still taking advantage of the cloud.

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS IoT ROS Extension

https://github.com/aws-robotics/aws-iot-bridge-example

AWS IoT CoreAWS IoT Greengrass

Edge

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Fleet Management

AWS IoT CoreAWS IoT Greengrass

Edge

AWS RoboMaker

Deploy

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

ArtificialIntelligence

MachineLearning

ReinforcementLearning

SupervisedLearning

UnsupervisedLearning

Reinforcement learning in the broader AI context

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Machine Learning Overview

SUPERVISED UNSUPERVISED REINFORCEMENT

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Reinforcement Learning in the real world

Reward positive behavior

Don’t reward negative behavior

The result!

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Reinforcement Learning use cases

AUTONOMOUS CARS FINANCIAL TRADING DATACENTER COOLINGFLEET LOGISTICS

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

What is Reinforcement Learning?

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

How does learning happen?VALUE FUNCTION

POLICY FUNCTION

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

RL Algorithms – Value Approximation

CHALLENGE Explore all (state, action, outcome) combinations in a real world race to create the value function

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

RL Algorithms – Value Approximation

CHALLENGE Explore all (state, action, outcome) combinations in a real world race to create the value function.

VERDICT Not possible

SOLUTION Approximate the value function using supervised learning. Use data from exploration to create a model capable of estimating the likely cumulative reward for all possible actions in a state.

The action with the highest cumulative reward,

is the best action.

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

RL Algorithms – Policy Approximation

ALTERNATIVE Try to directly map actions to states without using a value function.

HOW Use supervised learning to train a policy function model that returns the best action for any state.

How do we know this is a good action? Because it gave the highest reward. Keep improving our model in the direction of the highest reward.

FAMOUS EXAMPLE

Dota2 Challenge - OpenAI Five

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Reinforcement Learning Algorithms Compared

Value Approximation Policy Approximation

AdvantagesMore stable performance when it works, and tends to

converge on global optimum

Effective in continuous action spaces, can learn stochastic policies,

and faster convergence

DisadvantagesDifficult to converge if too many (state, action)

combinations, slower convergence in general, and

can’t learn stochastic properties

Typically converges to a local rather than global optimum, high

variance in estimating the gradient adversely affects stability, and

evaluating a policy is generally inefficient

Examples Q-Learning, Deep Q Network, Deep Double Q Network Policy Gradient, Proximal Policy Optimization (PPO)

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Amazon SageMaker

Amazon SageMaker is a fully-managed servicethat covers the entire machine learningworkflow to label and prepare your data, choose an algorithm, train the algorithm, tune and optimize it for deployment, makepredictions, and take action.

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Amazon SageMaker RL

Amazon SageMaker RL builds on top ofAmazon SageMaker, adding pre-packaged RL toolkits and making it easy to integrate anysimulation environment.

• Intel Coach

• Ray RLLib

• TensorFlow,…

• Support for Open Source and

Commercial Environments

Pre-Packaged RL Toolkits

• OpenAI Gym

• MXNet

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

AWS Cloud

NAT

gateway

VPC

Models

Simulation

video

Metrics

Simulation Architecture

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Gazebo

Gazebo is an open-source,

Robot simulation tool.

https://gazebosim.org

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

Interactions with Simulator

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

In Closing

Accelerate development time

with zero infrastructure provisioning

through AWS services for

Cloud Robotics.

• AWS Robomaker

• AWS IoT Core

• AWS IoT Greengrass

• Amazon SageMaker RL

Cloud Robotics on AWS

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMIT

GO BUILD!

Özkan Can, AWS - Senior Solutions Architect

@_ozzc

© 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.SUMMITSUMMIT © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.

top related