SlideShare a Scribd company logo
1 of 42
Download to read offline
AI Integration WG
ALMA, SustainML & ROS 2 use case
28/06/2022
Agenda
02 Application Aware, Life-Cycle Oriented Model-HW Co-Design
Framework for Sustainable, Energy Efficient ML Systems
SustainML
01 Human Centric Algebraic Machine Learning
ALMA
03 Serving models to ROS 2 robots, with Vulcanexus
environment
Enabling ML technologies for ROS 2
Agenda
02 Application Aware, Life-Cycle Oriented Model-HW Co-Design
Framework for Sustainable, Energy Efficient ML Systems
SustainML
01 Human Centric Algebraic Machine Learning
ALMA
03 Serving models to ROS 2 robots, with Vulcanexus
environment
Enabling ML technologies for ROS 2
ALMA Mission
To provide a new ML paradigm, known as
AML
● Easily understandable by users (no black-box)
● Ease of interaction (Human-AML interaction)
● Seamless integration with AML-IP
● Ensure long-term maintenance of AML environments
Spain
Portugal
Germany
France Finland
What is AML?
Symbolic AI:
● Description of the world using formulas
● Difficulty learning from data
● High transparency
Statistical Learning (including Neural Networks):
● Learning from data
● Difficulty using formal descriptions
● Usually opaque
Algebraic Machine Learning:
● Description of the world using formulas
● Learning from data
● Can combine data and formal descriptions
● High transparency
AML vs Traditional approaches to Machine Learning
What is AML?
Symbolic AI:
● Uses symbols
● Symbols represent real world objects
● Mostly uses discrete mathematics
● Symbols are permanent
Statistical Learning (including Neural Networks):
● Uses parameters
● Parameters can map to the world or to intermediate
internal descriptions
● Mostly uses continuous mathematics
● Parameters can change
Algebraic Machine Learning:
● Uses symbols
● Symbols can map to the world (constants) or to
intermediate internal descriptions (atoms)
● Uses discrete mathematics
● Can create new symbols
● Symbols can change
AML vs Traditional approaches to Machine Learning
What is AML?
Main Algebraic Machine Learning features
Less sensitive to
statistical features
of training data
No tradeoff between
memorization and
learning (no overfitting)
High mathematical
transparency
Large-scale
distributed learning
Interactive ML
What is AML?
Main Algebraic Machine Learning features
Less sensitive to
statistical features
of training data
High mathematical
transparency
Distributed ML
ecosystem
Interactive ML
No tradeoff between
memorization and
learning (no overfitting)
Less sensitive to
statistical features
of training data
High mathematical
transparency
Distributed ML
ecosystem
Interactive ML
No tradeoff between
memorization and
learning (no overfitting)
What is AML?
Main Algebraic Machine Learning features
What is AML?
Main Algebraic Machine Learning features
Less sensitive to
statistical features
of training data
No tradeoff between
memorization and
learning (no overfitting)
High mathematical
transparency
Distributed ML
ecosystem
Interactive ML
What is AML?
Main Algebraic Machine Learning features
Less sensitive to
statistical features
of training data
No tradeoff between
memorization and
learning (no overfitting)
High mathematical
transparency
Large-scale
distributed learning
Interactive ML
What is AML?
Main Algebraic Machine Learning features
Less sensitive to
statistical features
of training data
No tradeoff between
memorization and
learning (no overfitting)
High mathematical
transparency
Large-scale
distributed learning
Interactive ML
AML - Proven Cases
Handwritten
digit recognition
- Supervised learning (MNIST)
- Atoms: algebraic elements
resulting from learning
Classification of human
motion
- Learning from both formal knowledge
and data (OPPORTUNITY)
- Activity recognition.
Queens
completion problem
- Learning from formal knowledge
- Rules encoded in the algebra
- AML understands the game from
the beginning of learning process
Proven cases of Algebraic Machine Learning
AML - Ongoing case studies
Ongoing Algebraic Machine Learning applications
World models
Robot control and
path planning
Interaction with
gesture keyboard
- Gesture keyboard
- Confidence feedback interface
- Robot control and landscape
navigation
- Formal description of high level
real-world concepts
- Ethical aspects
ALMA contribution to ROS 2
Develop an open-source decentralized AI integrating platform compatible with ROS 2
● AI-IP for cloud and edge computing environments
● Extend AI-IP to work with robotic and constrained device platforms.
ROS 2 as the primary robotic platform
● Open-source experimentation tools for ROS 2 community
eProsima is betting on AML as a key disrupting AI technology for robotics
● Integration of AML into a publish-subscribe framework compatible with ROS 2
1
2
Decentralized AI Integrating Platform fully compatible with ROS 2
Agenda
01 Human Centric Algebraic Machine Learning
ALMA
03 Serving models to ROS 2 robots, with Vulcanexus
environment
Enabling ML technologies for ROS 2
02 Application Aware, Life-Cycle Oriented Model-HW Co-Design
Framework for Sustainable, Energy Efficient ML Systems
SustainML
SustainML Mission
To provide a new SUSTAINABLE and INTERACTIVE ML
framework development for Green AI that will:
● Comprehensively prioritize and advocate energy
efficiency across the entire lifecycle of an application
● Avoid AI-waste
Spain
Germany
France
Denmark
Switzerland
SustainML framework
How the SustainML framework will work?
Developers describe the
tasks to solve a specific
problem
Describe the problem Framework analyze the tasks,
divide and encode the problem into
an abstract functional semantic
Encode the tasks
SustainML
framework
interactive
process
Developers can reconfigure the model with
other cores with optional pretrained
parameters or design their own models
Interactive design process
Framework suggests ML models with knowledge
transfer and recycling from its collection of Neural
Networks knowledge cores
Model suggestion
The framework will convey top tier AI experts’
experience by suggesting best practices, more
efficient alternatives or avoiding previous
negative results
Framework understands the problem
Not yet another AI studio tool!
SustainML project will focus on:
● Investigate the detailed footprint on computing and data HW
● Develop novel HW accelerators optimized for different layers and operations
The user will see the estimated CO2 footprint, HW resource and training
time during design process (before these costs actually occur!).
Useful for any user:
● Novice: what functionality different parts of a NN carries and why certain model
structures are better suited for specific tasks
● Intermediate/Experts: leverage SustainML framework to develop efficient models
optimized for their goals.
● AI researchers: skip the problem describing process and use the framework to
benchmark and optimize their ML models’ carbon footprint
1
2
3
Agenda
02 Application Aware, Life-Cycle Oriented Model-HW Co-Design
Framework for Sustainable, Energy Efficient ML Systems
SustainML
01 Human Centric Algebraic Machine Learning
ALMA
03 Serving models to ROS 2 robots, with Vulcanexus
environment
Enabling ML technologies for ROS 2
Inference Server
21
LAN scenario
Inference Server
22
LAN scenario - Inference as a ROS 2 service
Inference service
23
Service Client and Server
Inference service
24
Service Client and Server
Inference service
25
Service Client and Server
Inference Server
26
LAN scenario
Inference Server
27
WAN scenario
eProsima DDS Router
28
Bridge DDS communications over WAN
eProsima DDS Router
29
Avoiding data introspection
achieves a zero-copy
system in data forwarding.
Efficient data routing
eProsima DDS Router features
Deploy intermediate nodes
to discover new entities
that enter and leave the
network dynamically.
Design for distributed
DDS networks
Supports WAN over TCP
communication to establish
DDS communications over
the Internet.
WAN communications
over TCP
Easily configurable modular
system for users with no
knowledge of computer
networks.
Easy deployment
Forward the user data
belonging to the specified
topics
Topic whitelisting
eProsima DDS Router
30
eProsima DDS Router for communication of edge devices with the Cloud.
Inference Server
31
WAN scenario
Inference Server
32
WAN scenario
Services support
currently under
development.
AI - Integrating Platform
Inference / Classification
Use The model to inference attributes over
the data received.
Train Model
Use the available data to find a better
model.
Publish Best Model
Publish the best model found via DDS topic.
Publish whole model or updates.
Install Best Model
Change the current model for the best
received.
Collect new labeled data to reinforce the
model.
AI
Integrating
Platform
Collect Data
Edge Side
Final Robot communicating via DDS.
Inference in edge.
Server Side
Great computation power.
Train new models.
33
Main features of AI - IP
Inference Server
34
WAN scenario (multiple servers)
AI - Integrating Platform
35
Distributed learning scenario
AI - Integrating Platform
36
Collaborative learning scenario
Vulcanexus
37
The All-in-One ROS 2 tool set
Easy to download and configure
Customize your experience with ROS 2
WAN communication out of the box
Fully synchronized with Fast-DDS updates
www.eProsima.com
Linkedin.com/company/eProsima Twitter.com/EProsima
Annex
Real-Time model update
Update to best model found
in the network in run-time
automatically
Parallel training
Multiple nodes could be
searching for the best
solution at the same time
Free Edge of training
Reduce edge device
computation effort to the
strictly needed
ML Model independent
Platform could be used
independently of the ML
model
Neural Networks, Random
Forest, AML, etc.
Collaborative network
Everyone gets benefited by
collaborating into finding the
best solution
Private dataset
Sharing the trained model
and not the dataset keeps
your data private and safe
40
Features and Capabilities
AI - Integrating Platform
Advantages of a ROS 2 AI Integration WG leaded by eProsima
Use the effort invested in ALMA and
its results to enhance the
integration of AI in ROS 2
Seamless use the AI Integrating
Platform in the ROS 2 ecosystem
Support for early use of AML in ROS
2
Discuss state-of-the-art topics with
experts on the field taking advantage
of ALMA partners
Promote contribution to
open-source AI integration
tools for ROS 2
Dissemination of ROS 2 by means of
ALMA conferences
ROS 2 AI Integration WG
41
Inference Server
42
WAN scenario (multiple servers)

More Related Content

What's hot

ROSCON Fr: is ROS 2 ready for production?
ROSCON Fr: is ROS 2 ready for production?ROSCON Fr: is ROS 2 ready for production?
ROSCON Fr: is ROS 2 ready for production?Thomas Moulard
 
An Introduction To Server Virtualisation
An Introduction To Server VirtualisationAn Introduction To Server Virtualisation
An Introduction To Server VirtualisationAlan McSweeney
 
VMware NSX 101: What, Why & How
VMware NSX 101: What, Why & HowVMware NSX 101: What, Why & How
VMware NSX 101: What, Why & HowAniekan Akpaffiong
 
ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020
ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020
ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020Joe Speed
 
Object Storage in a Cloud-Native Container Envirnoment
Object Storage in a Cloud-Native Container EnvirnomentObject Storage in a Cloud-Native Container Envirnoment
Object Storage in a Cloud-Native Container EnvirnomentMinio
 
【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪
【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪
【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪MAKERPRO.cc
 
Cloud computing
Cloud computingCloud computing
Cloud computingstudent
 
Curso de Performance and Tuning - Linux
Curso de Performance and Tuning - LinuxCurso de Performance and Tuning - Linux
Curso de Performance and Tuning - LinuxDell Technologies
 
Federated Cloud Computing - The OpenNebula Experience v1.0s
Federated Cloud Computing  - The OpenNebula Experience v1.0sFederated Cloud Computing  - The OpenNebula Experience v1.0s
Federated Cloud Computing - The OpenNebula Experience v1.0sIgnacio M. Llorente
 
Google Cloud Networking Deep Dive
Google Cloud Networking Deep DiveGoogle Cloud Networking Deep Dive
Google Cloud Networking Deep DiveMichelle Holley
 
NIST Cloud Computing Reference Architecture
NIST Cloud Computing Reference ArchitectureNIST Cloud Computing Reference Architecture
NIST Cloud Computing Reference ArchitectureThanakrit Lersmethasakul
 
"The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta...
"The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta..."The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta...
"The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta...Edge AI and Vision Alliance
 
Samsung: CMM-H Tiered Memory Solution with Built-in DRAM
Samsung: CMM-H Tiered Memory Solution with Built-in DRAMSamsung: CMM-H Tiered Memory Solution with Built-in DRAM
Samsung: CMM-H Tiered Memory Solution with Built-in DRAMMemory Fabric Forum
 
DDS: The IoT Data Sharing Standard
DDS: The IoT Data Sharing StandardDDS: The IoT Data Sharing Standard
DDS: The IoT Data Sharing StandardAngelo Corsaro
 
A tutorial on CloudSim
A tutorial on CloudSimA tutorial on CloudSim
A tutorial on CloudSimHabibur Rahman
 

What's hot (20)

ROSCON Fr: is ROS 2 ready for production?
ROSCON Fr: is ROS 2 ready for production?ROSCON Fr: is ROS 2 ready for production?
ROSCON Fr: is ROS 2 ready for production?
 
An Introduction To Server Virtualisation
An Introduction To Server VirtualisationAn Introduction To Server Virtualisation
An Introduction To Server Virtualisation
 
VMware NSX 101: What, Why & How
VMware NSX 101: What, Why & HowVMware NSX 101: What, Why & How
VMware NSX 101: What, Why & How
 
ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020
ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020
ROS 2 Foxy with Eclipse Cyclone DDS | Philly ROS Meetup July 20th 2020
 
Introduction to CloudStack
Introduction to CloudStack Introduction to CloudStack
Introduction to CloudStack
 
EIN overview
EIN overviewEIN overview
EIN overview
 
Object Storage in a Cloud-Native Container Envirnoment
Object Storage in a Cloud-Native Container EnvirnomentObject Storage in a Cloud-Native Container Envirnoment
Object Storage in a Cloud-Native Container Envirnoment
 
【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪
【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪
【1110ROS社群開講】ROS 2與DDS應用於工業領域_王健豪
 
Cloud computing
Cloud computingCloud computing
Cloud computing
 
Curso de Performance and Tuning - Linux
Curso de Performance and Tuning - LinuxCurso de Performance and Tuning - Linux
Curso de Performance and Tuning - Linux
 
Federated Cloud Computing - The OpenNebula Experience v1.0s
Federated Cloud Computing  - The OpenNebula Experience v1.0sFederated Cloud Computing  - The OpenNebula Experience v1.0s
Federated Cloud Computing - The OpenNebula Experience v1.0s
 
Google Cloud Networking Deep Dive
Google Cloud Networking Deep DiveGoogle Cloud Networking Deep Dive
Google Cloud Networking Deep Dive
 
NIST Cloud Computing Reference Architecture
NIST Cloud Computing Reference ArchitectureNIST Cloud Computing Reference Architecture
NIST Cloud Computing Reference Architecture
 
Cloud computing presentation
Cloud computing presentationCloud computing presentation
Cloud computing presentation
 
"The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta...
"The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta..."The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta...
"The Xilinx AI Engine: High Performance with Future-proof Architecture Adapta...
 
Samsung: CMM-H Tiered Memory Solution with Built-in DRAM
Samsung: CMM-H Tiered Memory Solution with Built-in DRAMSamsung: CMM-H Tiered Memory Solution with Built-in DRAM
Samsung: CMM-H Tiered Memory Solution with Built-in DRAM
 
DDS: The IoT Data Sharing Standard
DDS: The IoT Data Sharing StandardDDS: The IoT Data Sharing Standard
DDS: The IoT Data Sharing Standard
 
SOME_IP_29408.pdf
SOME_IP_29408.pdfSOME_IP_29408.pdf
SOME_IP_29408.pdf
 
A tutorial on CloudSim
A tutorial on CloudSimA tutorial on CloudSim
A tutorial on CloudSim
 
Udev
UdevUdev
Udev
 

Similar to ROS 2 AI Integration Working Group 1: ALMA, SustainML & ROS 2 use case

Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...
Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...
Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...IRJET Journal
 
Serverless machine learning architectures at Helixa
Serverless machine learning architectures at HelixaServerless machine learning architectures at Helixa
Serverless machine learning architectures at HelixaData Science Milan
 
xUMLFinalPresentation.ppt
xUMLFinalPresentation.pptxUMLFinalPresentation.ppt
xUMLFinalPresentation.pptssuser2ef938
 
Zeller Edm Summit Agile Deployment Of Predictive Analytics
Zeller Edm Summit   Agile Deployment Of Predictive AnalyticsZeller Edm Summit   Agile Deployment Of Predictive Analytics
Zeller Edm Summit Agile Deployment Of Predictive AnalyticsRonald.Ramos
 
Large Language Models Bootcamp
Large Language Models BootcampLarge Language Models Bootcamp
Large Language Models BootcampData Science Dojo
 
BSSML16 L10. Summary Day 2 Sessions
BSSML16 L10. Summary Day 2 SessionsBSSML16 L10. Summary Day 2 Sessions
BSSML16 L10. Summary Day 2 SessionsBigML, Inc
 
The road ahead for architectural languages [ACVI 2016]
The road ahead for architectural languages [ACVI 2016]The road ahead for architectural languages [ACVI 2016]
The road ahead for architectural languages [ACVI 2016]Ivano Malavolta
 
Qiang Yu Resume
Qiang Yu Resume Qiang Yu Resume
Qiang Yu Resume Qiang Yu
 
Qiang Yu CV
Qiang Yu CVQiang Yu CV
Qiang Yu CVQiang Yu
 
Simulink based software
Simulink based software Simulink based software
Simulink based software DorleControls
 
The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5
The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5
The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5Robert Grossman
 
The REMICS model-driven process for migrating legacy applications to the cloud
The REMICS model-driven process for migrating legacy applications to the cloudThe REMICS model-driven process for migrating legacy applications to the cloud
The REMICS model-driven process for migrating legacy applications to the cloudMarcos Almeida
 
MLFlow: Platform for Complete Machine Learning Lifecycle
MLFlow: Platform for Complete Machine Learning Lifecycle MLFlow: Platform for Complete Machine Learning Lifecycle
MLFlow: Platform for Complete Machine Learning Lifecycle Databricks
 
Building a MLOps Platform Around MLflow to Enable Model Productionalization i...
Building a MLOps Platform Around MLflow to Enable Model Productionalization i...Building a MLOps Platform Around MLflow to Enable Model Productionalization i...
Building a MLOps Platform Around MLflow to Enable Model Productionalization i...Databricks
 
Industrial Edge.pptx
Industrial Edge.pptxIndustrial Edge.pptx
Industrial Edge.pptxSripad NS
 
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...Srivatsan Ramanujam
 
IRJET - A Review on Chatbot Design and Implementation Techniques
IRJET -  	  A Review on Chatbot Design and Implementation TechniquesIRJET -  	  A Review on Chatbot Design and Implementation Techniques
IRJET - A Review on Chatbot Design and Implementation TechniquesIRJET Journal
 
Software Architecture in Process Automation: UML & the "Smart Factory"
Software Architecture in Process Automation: UML & the "Smart Factory"Software Architecture in Process Automation: UML & the "Smart Factory"
Software Architecture in Process Automation: UML & the "Smart Factory"Heiko Koziolek
 

Similar to ROS 2 AI Integration Working Group 1: ALMA, SustainML & ROS 2 use case (20)

Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...
Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...
Limited Budget but Effective End to End MLOps Practices (Machine Learning Mod...
 
Serverless machine learning architectures at Helixa
Serverless machine learning architectures at HelixaServerless machine learning architectures at Helixa
Serverless machine learning architectures at Helixa
 
xUMLFinalPresentation.ppt
xUMLFinalPresentation.pptxUMLFinalPresentation.ppt
xUMLFinalPresentation.ppt
 
Zeller Edm Summit Agile Deployment Of Predictive Analytics
Zeller Edm Summit   Agile Deployment Of Predictive AnalyticsZeller Edm Summit   Agile Deployment Of Predictive Analytics
Zeller Edm Summit Agile Deployment Of Predictive Analytics
 
Large Language Models Bootcamp
Large Language Models BootcampLarge Language Models Bootcamp
Large Language Models Bootcamp
 
BSSML16 L10. Summary Day 2 Sessions
BSSML16 L10. Summary Day 2 SessionsBSSML16 L10. Summary Day 2 Sessions
BSSML16 L10. Summary Day 2 Sessions
 
O Matrix Overview
O Matrix OverviewO Matrix Overview
O Matrix Overview
 
The road ahead for architectural languages [ACVI 2016]
The road ahead for architectural languages [ACVI 2016]The road ahead for architectural languages [ACVI 2016]
The road ahead for architectural languages [ACVI 2016]
 
Qiang Yu Resume
Qiang Yu Resume Qiang Yu Resume
Qiang Yu Resume
 
Qiang Yu CV
Qiang Yu CVQiang Yu CV
Qiang Yu CV
 
Introduction to ML.NET
Introduction to ML.NETIntroduction to ML.NET
Introduction to ML.NET
 
Simulink based software
Simulink based software Simulink based software
Simulink based software
 
The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5
The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5
The Impact of Cloud Computing on Predictive Analytics 7-29-09 v5
 
The REMICS model-driven process for migrating legacy applications to the cloud
The REMICS model-driven process for migrating legacy applications to the cloudThe REMICS model-driven process for migrating legacy applications to the cloud
The REMICS model-driven process for migrating legacy applications to the cloud
 
MLFlow: Platform for Complete Machine Learning Lifecycle
MLFlow: Platform for Complete Machine Learning Lifecycle MLFlow: Platform for Complete Machine Learning Lifecycle
MLFlow: Platform for Complete Machine Learning Lifecycle
 
Building a MLOps Platform Around MLflow to Enable Model Productionalization i...
Building a MLOps Platform Around MLflow to Enable Model Productionalization i...Building a MLOps Platform Around MLflow to Enable Model Productionalization i...
Building a MLOps Platform Around MLflow to Enable Model Productionalization i...
 
Industrial Edge.pptx
Industrial Edge.pptxIndustrial Edge.pptx
Industrial Edge.pptx
 
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
A Pipeline for Distributed Topic and Sentiment Analysis of Tweets on Pivotal ...
 
IRJET - A Review on Chatbot Design and Implementation Techniques
IRJET -  	  A Review on Chatbot Design and Implementation TechniquesIRJET -  	  A Review on Chatbot Design and Implementation Techniques
IRJET - A Review on Chatbot Design and Implementation Techniques
 
Software Architecture in Process Automation: UML & the "Smart Factory"
Software Architecture in Process Automation: UML & the "Smart Factory"Software Architecture in Process Automation: UML & the "Smart Factory"
Software Architecture in Process Automation: UML & the "Smart Factory"
 

More from eProsima

AML & ALMA: Project Overview
AML & ALMA: Project OverviewAML & ALMA: Project Overview
AML & ALMA: Project OvervieweProsima
 
micro-ROS - ROS 2 into microcontrollers
micro-ROS - ROS 2 into microcontrollersmicro-ROS - ROS 2 into microcontrollers
micro-ROS - ROS 2 into microcontrollerseProsima
 
Fast DDS Hello World in Windows
Fast DDS Hello World in WindowsFast DDS Hello World in Windows
Fast DDS Hello World in WindowseProsima
 
ROS 2 deployment in K8s: DDS Router as WAN comms enabler
ROS 2 deployment in K8s: DDS Router as WAN comms enablerROS 2 deployment in K8s: DDS Router as WAN comms enabler
ROS 2 deployment in K8s: DDS Router as WAN comms enablereProsima
 
Algebraic Machine Learning - On changing the rules of the game
Algebraic Machine Learning - On changing the rules of the gameAlgebraic Machine Learning - On changing the rules of the game
Algebraic Machine Learning - On changing the rules of the gameeProsima
 
Vulcanexus: the all-in-one ROS 2 toolset
Vulcanexus: the all-in-one ROS 2 toolsetVulcanexus: the all-in-one ROS 2 toolset
Vulcanexus: the all-in-one ROS 2 toolseteProsima
 
micro-ROS Humble Hawskbill
micro-ROS Humble Hawskbillmicro-ROS Humble Hawskbill
micro-ROS Humble HawskbilleProsima
 
eProsima DDS Router: a micro-ROS Cloud scenario
eProsima DDS Router: a micro-ROS Cloud scenarioeProsima DDS Router: a micro-ROS Cloud scenario
eProsima DDS Router: a micro-ROS Cloud scenarioeProsima
 
ALMA - Integration of AI in ROS 2 ecosystem
ALMA - Integration of AI in ROS 2 ecosystemALMA - Integration of AI in ROS 2 ecosystem
ALMA - Integration of AI in ROS 2 ecosystemeProsima
 
micro-ROS: Developing ROS 2 professional applications based on MCUs
micro-ROS: Developing ROS 2 professional applications based on MCUsmicro-ROS: Developing ROS 2 professional applications based on MCUs
micro-ROS: Developing ROS 2 professional applications based on MCUseProsima
 
micro-ROS goes easy: Developing professional applications using Eclipse based...
micro-ROS goes easy: Developing professional applications using Eclipse based...micro-ROS goes easy: Developing professional applications using Eclipse based...
micro-ROS goes easy: Developing professional applications using Eclipse based...eProsima
 
micro-ROS - New client library and middleware features
micro-ROS - New client library and middleware featuresmicro-ROS - New client library and middleware features
micro-ROS - New client library and middleware featureseProsima
 
Towards Easy 5GS Integration in ROS2 - eProsima & Ericsson
Towards Easy 5GS Integration in ROS2 - eProsima & EricssonTowards Easy 5GS Integration in ROS2 - eProsima & Ericsson
Towards Easy 5GS Integration in ROS2 - eProsima & EricssoneProsima
 
Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...
Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...
Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...eProsima
 
Micro XRCE-DDS and micro-ROS
Micro XRCE-DDS and micro-ROSMicro XRCE-DDS and micro-ROS
Micro XRCE-DDS and micro-ROSeProsima
 
eProsima - Company brief
eProsima - Company briefeProsima - Company brief
eProsima - Company briefeProsima
 
Fast DDS Features & Tools
Fast DDS Features & ToolsFast DDS Features & Tools
Fast DDS Features & ToolseProsima
 
micro-ROS Galactic
micro-ROS Galacticmicro-ROS Galactic
micro-ROS GalacticeProsima
 
Integration Service: Integrating Communication Protocols
Integration Service: Integrating Communication ProtocolsIntegration Service: Integrating Communication Protocols
Integration Service: Integrating Communication ProtocolseProsima
 
Micro XRCE-DDS: Bringing DDS into microcontrollers
Micro XRCE-DDS: Bringing DDS into microcontrollersMicro XRCE-DDS: Bringing DDS into microcontrollers
Micro XRCE-DDS: Bringing DDS into microcontrollerseProsima
 

More from eProsima (20)

AML & ALMA: Project Overview
AML & ALMA: Project OverviewAML & ALMA: Project Overview
AML & ALMA: Project Overview
 
micro-ROS - ROS 2 into microcontrollers
micro-ROS - ROS 2 into microcontrollersmicro-ROS - ROS 2 into microcontrollers
micro-ROS - ROS 2 into microcontrollers
 
Fast DDS Hello World in Windows
Fast DDS Hello World in WindowsFast DDS Hello World in Windows
Fast DDS Hello World in Windows
 
ROS 2 deployment in K8s: DDS Router as WAN comms enabler
ROS 2 deployment in K8s: DDS Router as WAN comms enablerROS 2 deployment in K8s: DDS Router as WAN comms enabler
ROS 2 deployment in K8s: DDS Router as WAN comms enabler
 
Algebraic Machine Learning - On changing the rules of the game
Algebraic Machine Learning - On changing the rules of the gameAlgebraic Machine Learning - On changing the rules of the game
Algebraic Machine Learning - On changing the rules of the game
 
Vulcanexus: the all-in-one ROS 2 toolset
Vulcanexus: the all-in-one ROS 2 toolsetVulcanexus: the all-in-one ROS 2 toolset
Vulcanexus: the all-in-one ROS 2 toolset
 
micro-ROS Humble Hawskbill
micro-ROS Humble Hawskbillmicro-ROS Humble Hawskbill
micro-ROS Humble Hawskbill
 
eProsima DDS Router: a micro-ROS Cloud scenario
eProsima DDS Router: a micro-ROS Cloud scenarioeProsima DDS Router: a micro-ROS Cloud scenario
eProsima DDS Router: a micro-ROS Cloud scenario
 
ALMA - Integration of AI in ROS 2 ecosystem
ALMA - Integration of AI in ROS 2 ecosystemALMA - Integration of AI in ROS 2 ecosystem
ALMA - Integration of AI in ROS 2 ecosystem
 
micro-ROS: Developing ROS 2 professional applications based on MCUs
micro-ROS: Developing ROS 2 professional applications based on MCUsmicro-ROS: Developing ROS 2 professional applications based on MCUs
micro-ROS: Developing ROS 2 professional applications based on MCUs
 
micro-ROS goes easy: Developing professional applications using Eclipse based...
micro-ROS goes easy: Developing professional applications using Eclipse based...micro-ROS goes easy: Developing professional applications using Eclipse based...
micro-ROS goes easy: Developing professional applications using Eclipse based...
 
micro-ROS - New client library and middleware features
micro-ROS - New client library and middleware featuresmicro-ROS - New client library and middleware features
micro-ROS - New client library and middleware features
 
Towards Easy 5GS Integration in ROS2 - eProsima & Ericsson
Towards Easy 5GS Integration in ROS2 - eProsima & EricssonTowards Easy 5GS Integration in ROS2 - eProsima & Ericsson
Towards Easy 5GS Integration in ROS2 - eProsima & Ericsson
 
Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...
Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...
Open Middleware Technologies for Smart Robotics - a FIWARE Smart Fest present...
 
Micro XRCE-DDS and micro-ROS
Micro XRCE-DDS and micro-ROSMicro XRCE-DDS and micro-ROS
Micro XRCE-DDS and micro-ROS
 
eProsima - Company brief
eProsima - Company briefeProsima - Company brief
eProsima - Company brief
 
Fast DDS Features & Tools
Fast DDS Features & ToolsFast DDS Features & Tools
Fast DDS Features & Tools
 
micro-ROS Galactic
micro-ROS Galacticmicro-ROS Galactic
micro-ROS Galactic
 
Integration Service: Integrating Communication Protocols
Integration Service: Integrating Communication ProtocolsIntegration Service: Integrating Communication Protocols
Integration Service: Integrating Communication Protocols
 
Micro XRCE-DDS: Bringing DDS into microcontrollers
Micro XRCE-DDS: Bringing DDS into microcontrollersMicro XRCE-DDS: Bringing DDS into microcontrollers
Micro XRCE-DDS: Bringing DDS into microcontrollers
 

Recently uploaded

"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks..."LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...Fwdays
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationSafe Software
 
Understanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitectureUnderstanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitecturePixlogix Infotech
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 3652toLead Limited
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii SoldatenkoFwdays
 
Unraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfUnraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfAlex Barbosa Coqueiro
 
Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Mattias Andersson
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationSlibray Presentation
 
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024BookNet Canada
 
SAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptxSAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptxNavinnSomaal
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Patryk Bandurski
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024Scott Keck-Warren
 
APIForce Zurich 5 April Automation LPDG
APIForce Zurich 5 April  Automation LPDGAPIForce Zurich 5 April  Automation LPDG
APIForce Zurich 5 April Automation LPDGMarianaLemus7
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machinePadma Pradeep
 
Powerpoint exploring the locations used in television show Time Clash
Powerpoint exploring the locations used in television show Time ClashPowerpoint exploring the locations used in television show Time Clash
Powerpoint exploring the locations used in television show Time Clashcharlottematthew16
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsRizwan Syed
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupFlorian Wilhelm
 
Developer Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLDeveloper Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLScyllaDB
 

Recently uploaded (20)

Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort ServiceHot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
 
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks..."LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
 
Understanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitectureUnderstanding the Laravel MVC Architecture
Understanding the Laravel MVC Architecture
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko
 
Unraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfUnraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdf
 
Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck Presentation
 
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC CataList - Tech Forum 2024
 
SAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptxSAP Build Work Zone - Overview L2-L3.pptx
SAP Build Work Zone - Overview L2-L3.pptx
 
E-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptx
E-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptxE-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptx
E-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptx
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024
 
APIForce Zurich 5 April Automation LPDG
APIForce Zurich 5 April  Automation LPDGAPIForce Zurich 5 April  Automation LPDG
APIForce Zurich 5 April Automation LPDG
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machine
 
Powerpoint exploring the locations used in television show Time Clash
Powerpoint exploring the locations used in television show Time ClashPowerpoint exploring the locations used in television show Time Clash
Powerpoint exploring the locations used in television show Time Clash
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL Certs
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project Setup
 
Developer Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLDeveloper Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQL
 

ROS 2 AI Integration Working Group 1: ALMA, SustainML & ROS 2 use case

  • 1. AI Integration WG ALMA, SustainML & ROS 2 use case 28/06/2022
  • 2. Agenda 02 Application Aware, Life-Cycle Oriented Model-HW Co-Design Framework for Sustainable, Energy Efficient ML Systems SustainML 01 Human Centric Algebraic Machine Learning ALMA 03 Serving models to ROS 2 robots, with Vulcanexus environment Enabling ML technologies for ROS 2
  • 3. Agenda 02 Application Aware, Life-Cycle Oriented Model-HW Co-Design Framework for Sustainable, Energy Efficient ML Systems SustainML 01 Human Centric Algebraic Machine Learning ALMA 03 Serving models to ROS 2 robots, with Vulcanexus environment Enabling ML technologies for ROS 2
  • 4. ALMA Mission To provide a new ML paradigm, known as AML ● Easily understandable by users (no black-box) ● Ease of interaction (Human-AML interaction) ● Seamless integration with AML-IP ● Ensure long-term maintenance of AML environments Spain Portugal Germany France Finland
  • 5. What is AML? Symbolic AI: ● Description of the world using formulas ● Difficulty learning from data ● High transparency Statistical Learning (including Neural Networks): ● Learning from data ● Difficulty using formal descriptions ● Usually opaque Algebraic Machine Learning: ● Description of the world using formulas ● Learning from data ● Can combine data and formal descriptions ● High transparency AML vs Traditional approaches to Machine Learning
  • 6. What is AML? Symbolic AI: ● Uses symbols ● Symbols represent real world objects ● Mostly uses discrete mathematics ● Symbols are permanent Statistical Learning (including Neural Networks): ● Uses parameters ● Parameters can map to the world or to intermediate internal descriptions ● Mostly uses continuous mathematics ● Parameters can change Algebraic Machine Learning: ● Uses symbols ● Symbols can map to the world (constants) or to intermediate internal descriptions (atoms) ● Uses discrete mathematics ● Can create new symbols ● Symbols can change AML vs Traditional approaches to Machine Learning
  • 7. What is AML? Main Algebraic Machine Learning features Less sensitive to statistical features of training data No tradeoff between memorization and learning (no overfitting) High mathematical transparency Large-scale distributed learning Interactive ML
  • 8. What is AML? Main Algebraic Machine Learning features Less sensitive to statistical features of training data High mathematical transparency Distributed ML ecosystem Interactive ML No tradeoff between memorization and learning (no overfitting)
  • 9. Less sensitive to statistical features of training data High mathematical transparency Distributed ML ecosystem Interactive ML No tradeoff between memorization and learning (no overfitting) What is AML? Main Algebraic Machine Learning features
  • 10. What is AML? Main Algebraic Machine Learning features Less sensitive to statistical features of training data No tradeoff between memorization and learning (no overfitting) High mathematical transparency Distributed ML ecosystem Interactive ML
  • 11. What is AML? Main Algebraic Machine Learning features Less sensitive to statistical features of training data No tradeoff between memorization and learning (no overfitting) High mathematical transparency Large-scale distributed learning Interactive ML
  • 12. What is AML? Main Algebraic Machine Learning features Less sensitive to statistical features of training data No tradeoff between memorization and learning (no overfitting) High mathematical transparency Large-scale distributed learning Interactive ML
  • 13. AML - Proven Cases Handwritten digit recognition - Supervised learning (MNIST) - Atoms: algebraic elements resulting from learning Classification of human motion - Learning from both formal knowledge and data (OPPORTUNITY) - Activity recognition. Queens completion problem - Learning from formal knowledge - Rules encoded in the algebra - AML understands the game from the beginning of learning process Proven cases of Algebraic Machine Learning
  • 14. AML - Ongoing case studies Ongoing Algebraic Machine Learning applications World models Robot control and path planning Interaction with gesture keyboard - Gesture keyboard - Confidence feedback interface - Robot control and landscape navigation - Formal description of high level real-world concepts - Ethical aspects
  • 15. ALMA contribution to ROS 2 Develop an open-source decentralized AI integrating platform compatible with ROS 2 ● AI-IP for cloud and edge computing environments ● Extend AI-IP to work with robotic and constrained device platforms. ROS 2 as the primary robotic platform ● Open-source experimentation tools for ROS 2 community eProsima is betting on AML as a key disrupting AI technology for robotics ● Integration of AML into a publish-subscribe framework compatible with ROS 2 1 2 Decentralized AI Integrating Platform fully compatible with ROS 2
  • 16. Agenda 01 Human Centric Algebraic Machine Learning ALMA 03 Serving models to ROS 2 robots, with Vulcanexus environment Enabling ML technologies for ROS 2 02 Application Aware, Life-Cycle Oriented Model-HW Co-Design Framework for Sustainable, Energy Efficient ML Systems SustainML
  • 17. SustainML Mission To provide a new SUSTAINABLE and INTERACTIVE ML framework development for Green AI that will: ● Comprehensively prioritize and advocate energy efficiency across the entire lifecycle of an application ● Avoid AI-waste Spain Germany France Denmark Switzerland
  • 18. SustainML framework How the SustainML framework will work? Developers describe the tasks to solve a specific problem Describe the problem Framework analyze the tasks, divide and encode the problem into an abstract functional semantic Encode the tasks SustainML framework interactive process Developers can reconfigure the model with other cores with optional pretrained parameters or design their own models Interactive design process Framework suggests ML models with knowledge transfer and recycling from its collection of Neural Networks knowledge cores Model suggestion The framework will convey top tier AI experts’ experience by suggesting best practices, more efficient alternatives or avoiding previous negative results Framework understands the problem
  • 19. Not yet another AI studio tool! SustainML project will focus on: ● Investigate the detailed footprint on computing and data HW ● Develop novel HW accelerators optimized for different layers and operations The user will see the estimated CO2 footprint, HW resource and training time during design process (before these costs actually occur!). Useful for any user: ● Novice: what functionality different parts of a NN carries and why certain model structures are better suited for specific tasks ● Intermediate/Experts: leverage SustainML framework to develop efficient models optimized for their goals. ● AI researchers: skip the problem describing process and use the framework to benchmark and optimize their ML models’ carbon footprint 1 2 3
  • 20. Agenda 02 Application Aware, Life-Cycle Oriented Model-HW Co-Design Framework for Sustainable, Energy Efficient ML Systems SustainML 01 Human Centric Algebraic Machine Learning ALMA 03 Serving models to ROS 2 robots, with Vulcanexus environment Enabling ML technologies for ROS 2
  • 22. Inference Server 22 LAN scenario - Inference as a ROS 2 service
  • 28. eProsima DDS Router 28 Bridge DDS communications over WAN
  • 29. eProsima DDS Router 29 Avoiding data introspection achieves a zero-copy system in data forwarding. Efficient data routing eProsima DDS Router features Deploy intermediate nodes to discover new entities that enter and leave the network dynamically. Design for distributed DDS networks Supports WAN over TCP communication to establish DDS communications over the Internet. WAN communications over TCP Easily configurable modular system for users with no knowledge of computer networks. Easy deployment Forward the user data belonging to the specified topics Topic whitelisting
  • 30. eProsima DDS Router 30 eProsima DDS Router for communication of edge devices with the Cloud.
  • 32. Inference Server 32 WAN scenario Services support currently under development.
  • 33. AI - Integrating Platform Inference / Classification Use The model to inference attributes over the data received. Train Model Use the available data to find a better model. Publish Best Model Publish the best model found via DDS topic. Publish whole model or updates. Install Best Model Change the current model for the best received. Collect new labeled data to reinforce the model. AI Integrating Platform Collect Data Edge Side Final Robot communicating via DDS. Inference in edge. Server Side Great computation power. Train new models. 33 Main features of AI - IP
  • 34. Inference Server 34 WAN scenario (multiple servers)
  • 35. AI - Integrating Platform 35 Distributed learning scenario
  • 36. AI - Integrating Platform 36 Collaborative learning scenario
  • 37. Vulcanexus 37 The All-in-One ROS 2 tool set Easy to download and configure Customize your experience with ROS 2 WAN communication out of the box Fully synchronized with Fast-DDS updates
  • 39. Annex
  • 40. Real-Time model update Update to best model found in the network in run-time automatically Parallel training Multiple nodes could be searching for the best solution at the same time Free Edge of training Reduce edge device computation effort to the strictly needed ML Model independent Platform could be used independently of the ML model Neural Networks, Random Forest, AML, etc. Collaborative network Everyone gets benefited by collaborating into finding the best solution Private dataset Sharing the trained model and not the dataset keeps your data private and safe 40 Features and Capabilities AI - Integrating Platform
  • 41. Advantages of a ROS 2 AI Integration WG leaded by eProsima Use the effort invested in ALMA and its results to enhance the integration of AI in ROS 2 Seamless use the AI Integrating Platform in the ROS 2 ecosystem Support for early use of AML in ROS 2 Discuss state-of-the-art topics with experts on the field taking advantage of ALMA partners Promote contribution to open-source AI integration tools for ROS 2 Dissemination of ROS 2 by means of ALMA conferences ROS 2 AI Integration WG 41
  • 42. Inference Server 42 WAN scenario (multiple servers)