You are on page 1of 36

Connectivity Standards in

Edge Autonomy Applications

Fernando Garcia-Aranda
& Gerardo Pardo-Castellote, Ph.D. IEEE Conference on Standards for
Communications and Networking
Real-Time Innovations, Inc.
October 29, 2019—Granada, Spain
©2019 Real-Time Innovations, Inc.
Edge Intelligence and Edge Autonomy

©2019 Real-Time Innovations, Inc.


Evolution of Distributed Computing…

Mainframes Client-Server Mobile Devices


Centralized Computing Distributed Computing Cloud Computing
1960s – 1970s 1980s – 2000s 2000s – 2010s

©2019 Real-Time Innovations, Inc.


Limitations of Centralized Models
• Centralized:
– Cameras send raw video and LIDAR to cloud
– Cloud processes, sends command to steer and break to car

Raw Data
STOP!

©2019 Real-Time Innovations, Inc.


Evolution of Distributed Computing…

Mainframes Client-Server Mobile Devices Edge Autonomy


Centralized Computing Distributed Computing Cloud Computing Edge Computing
1960s – 1970s 1980s – 2000s 2010s 2020s

©2019 Real-Time Innovations, Inc.


Leveraging Edge Autonomy
• Centralized:
• Edge Autonomy:
– Local sensor processing. Detect obstacles, road, ...
– Car makes decisions! Turn, accelerate, stop!
– Car to Car communication
– Cloud stores, analyzes, learns
TURN!

Status info

©2019 Real-Time Innovations, Inc.


Computational Challenges
2007 IBM Blue Gene/L, 400 TFOPS

40 TFLOPS Data Center circa 2002 (NEC Earth Simulation Computer)

2019 Nvidia
DGX-2 Station
2000 TFLOPS

©2019 Real-Time Innovations, Inc.


Data Distribution Challenges
Global M2M Traffic Growth
30

25 400,000 X Printed Material in Library of Congress


1,000 X Audio + Video in Library of Congress
20
Exabytes
per 15
Month
10

0
2017 2018 2019 2020 2021 2022

Source: Cisco VNI Global IP Traffic Forecast. 2017 – 2022


©2019 Real-Time Innovations, Inc.
IIoT Connectivity Frameworks

©2019 Real-Time Innovations, Inc.


IIoT Connectivity Stack
Participant X Participant Y
Semantic
Information Distributed Data Distributed Data
Interoperability and Management
Information (Data in Context) Interoperability
Interoperability and Management
(data context)

Syntactic
Framework Data (State, Events, Streams) Framework Interoperability
(data structures)
Connectivity
Technical
IICF Focus Transport Messages Transport Interoperability
(bytes)

Network Packets Network


Networking

Link Frames Link

Physical Bits Physical

©2019 Real-Time Innovations, Inc.


IIoT Connectivity Framework Functions
Semantic
Distributed Data Interoperability & Management Interoperability
(data context)

Framework

API Governance

Connectivity Syntactic
Publish-Subscribe Request-Reply Discovery Exception Handling
Framework Interoperability
Functions (data structures)
Quality
Data Resource Model
of Security
Service

State
Id and Addressing Data Type System Lifecycle (CRUD)
Management

Transport

Network

Link

Physical
©2019 Real-Time Innovations, Inc.
IIoT Core Connectivity Standards
Energy &
… Healthcare Manufacturing Transportation …
Utilities
Distributed Data Telecommunications Origin Manufacturing Origin
Interoperability and Management

DDS oneM2M Web OPC UA


Framework
Services

DDSI-RTPS CoAP MQTT HTTP OPC UA Bin


Transport

TCP UDP TCP

Network Internet Protocol (IP)

Link TSN / Wireless


Wireless Wireless Wireless
Ethernet LAN
PAN 2G/3G/LTE Wide Area
(802.1, (802.11 Wi-
(802.15) (3GPP) (802.16)
Physical 802.3) Fi)

©2019 Real-Time Innovations, Inc.


IIoT Core Connectivity Standards
Core Standard Criterion DDS Web Services OPC UA oneM2M

1 Provide syntactic interoperability Need XML or


✔ ✔ ✔
JSON
2 Open standard with strong independent, international governance ✔ ✔ ✔ ✔

3 Horizontal and neutral in its applicability across industries ✔ ✔ ✔ ✔

4 Stable and proven across multiple vertical industries Software Integration &
✔ Manufacturing Smart City Pilots*
Autonomy
5 Have standards-defined Core Gateways to all other core connectivity
standards Web Services, DDS, OPC UA, Web Services, Web Services,
OPC UA, oneM2M* oneM2M DDS, oneM2M* DDS*

6 Meet the connectivity framework functional requirements Pub-Sub in


✔ ✗ development

7 Meet non-functional requirements of performance, scalability, reliability, Reports not yet


Real-time in
resilience ✔ ✗ development
documented or
public
8 Meet security and safety requirements
✔ ✔ ✔ ✔

9 Not require any single component from any single vendor ✔ ✔ ✔ ✔

10 Have readily-available SDKs both commercial and open source ✔ ✔ ✔ ✔

©2019 Real-Time Innovations, Inc.


DDS Databus Databus

The Foundation for Edge Autonomy

©2019 Real-Time Innovations, Inc.


DDS Family of Specifications
Application
DDS-C++ DDS-JAVA DDS-IDL-C DDS-IDL-C#

DDS-SECURITY
DDS-OPC UA

DDS-XTYPES
DDS-JSON
DDS-WEB

DDS-XML
DDS-RPC
DDS-XRCE
IDL 4.
DDS v 1.4

RTPS v2.3
OPC/
HTTP
TCP

TCP UDP DTLS TLS


TSN
IP SHARED-MEMORY

Ethernet
©2019 Real-Time Innovations, Inc.
DDS Data-Centric Communication Model

Shared Global Dataspace (Domain)

Source (Key) Speed Power Position


CAR1 37.4 122.0 (37.41, -122.01)
CAR2 10.7 74.0 (36.95, -122.05)
CAR3 50.2 150.07 (37.42, -122.17)

QoS QoS
Topic D Topic D
DDS DATABUS

Persistence Recording
Service Service
©2019 Real-Time Innovations, Inc.
DDS Data-Centric Publish-Subscribe Model

©2019 Real-Time Innovations, Inc.


DDS Quality of Service (QoS) Policies
QoS Policy QoS Policy

User QoS
DURABILITY USER DATA
Cache

HISTORY TOPIC DATA

LIFESPAN GROUP DATA

Presentation
WRITER DATA LIFECYCLE PARTITION
Resources

READER DATA LIFECYCLE PRESENTATION

ENTITY FACTORY DESTINATION ORDER

Availability
RESOURCE LIMITS OWNERSHIP

RELIABILITY OWNERSHIP STRENGTH


Delivery

Transport
TIME BASED FILTER LIVELINESS

DEADLINE LATENCY BUDGET

CONTENT FILTERS TRANSPORT PRIORITY

©2019 Real-Time Innovations, Inc.


DDS RTPS Wire Protocol—Optimized for Real-Time
• Full peer-to-peer protocol
– No required brokers or servers
• Adaptable via QoS
– Reliability, timeouts, message priority RTPS
• Native reliable multicast support
– Uses transport multicast, if available, else unicast UDP
• Robust to disconnects
– Maintains session above (UDP) transport
• Efficient data serialization
– Binary XCDR
• Built-in availability and durability
– Durable & Persistent data, Historical cache, Failover support

©2019 Real-Time Innovations, Inc.


DDS Extensible Type System
@mutable
• Language-independent type system struct ShapeType {
@key string color;
– Language mappings to C, C++, Java, Ada, etc. @range(0, 200) long x;
@range(0, 250) long y;
• Type system can evolve and maintain };
@optional @min(5) float size;

interoperability struct ShapeTypeExt : ShapeType {


– Supports type inheritance @unit(“meter”) long x;

– Supports optional member


};

– Defines three levels of extensibility: final, /* Service definition */


enum Command { START, STOP };
appendable, and mutable
@service
• Dynamic APIs to access data and types interface RobotControl
{
– Systems can discover type definitions and create void command(Command com);
float setSpeed(float speed)
matching subscriber applications at runtime raises (TooFast);
float getSpeed();
};

©2019 Real-Time Innovations, Inc.


DDS Security Specification
• DDS Security specification
specifies mechanisms for:
– Authentication
– Access Control
– Data Integrity and
Confidentiality
– Enforce non-repudiation

©2019 Real-Time Innovations, Inc.


DDS-Based Integration

IIoT System
Mobile and Web User Interfaces Manufacturing Workcells

Light-Weight Sleep/Wake Clients


Clients Equipment
Autonomy

Connectivity Core Connectivity Core


Standard (HTTP/REST) Standard (OPC-UA)
Real-Time
Decisions
DDS-WEB OPC UA/DDS
Gateway Gateway

DDS Databus

Gateway

©2019 Real-Time Innovations, Inc.


DDS Databus—Data Exchange from Edge to Cloud

• End-User Applications
• Connect with RESTful/Web Sockets Public Internet
Cloud Databus
• Teleoperation
• Traffic Management

Unified Data Model


• Traffic Light Control
• Congestion Management Fleet Management Site Databus
• Road Management
• Environmental & Road Conditions

• Probe Data Collection (Sensors)


• Environmental Weather Unit Databus
V2V & V2X
• Dynamic Vehicle Location

• Collision Avoidance Think HMI

• Navigation In-Car Platform Machine Databus


• Safety Certified
Sense Act

©2019 Real-Time Innovations, Inc.


Example: Clinical Decision System Architecture

Admin Domain Bus (Cloud)

DDS
Router Gateway, IX, Enterprise, 3rd Party

Central Domain Bus

DDS DDS
Workstations, Storage,
Router Router
Historian
Room Domain Bus Room Domain Bus (Edge)

Patient Monitoring Devices


©2019 Real-Time Innovations, Inc.
Example: Gran Coulee Dam
Interested in
many quantities
Radar
Control
Radar Room Existing SCADA
Displays Logging (to be replaced)

Alarming Migration
Monitor Server

Control Room Bus

TCP (WAN)
Redundant
Routing
Local quantity
VPN/Firewall VPN/Firewall VPN/Firewall interest

Segment Bus Segment Bus Segment Bus

IPC IPC IPC


IPC IPC IPC
IPC IPC IPC

©2019 Real-Time Innovations, Inc.


Example: Autonomous Vehicles
Traffic Maps

CONNEXT DDS DATABUS

Situation
Situation Vehicle Error
Awareness Planning Logging
Awareness Control Management

CONNEXT DDS DATABUS

Sensing

Cameras, Vehicle
Cameras, Cameras,
Data Fusion Cameras,
Localization Visualization Navigation
LIDAR, Radar… Platform
LIDAR, Radar… LIDAR, Radar… LIDAR, Radar…

©2019 Real-Time Innovations, Inc.


DDS and TSN
A sneak peek into the next DDS standards…

©2019 Real-Time Innovations, Inc.


Time Sensitive Networking
• New set of IEEE standards
• Deterministic packet delivery on
CUC
standard Ethernet
User/Network
End Station User Configuration
Configuration Interface (UNI)
Protocol
CNC Network
Management
Protocol

Stream Stream
Talker End Stations Data Bridges Data Listener End Stations

©2019 Real-Time Innovations, Inc.


DDS + TSN
DW

A1 DW 1011 DR

DR 1001 DR

4001 DW A3

DR 2011 DR

A2 DW 2021 DR

DW
T period 1ms

T period 4ms

©2019 Real-Time Innovations, Inc.


DDS-TSN
• DDS and TSN are a good fit:
DDS ❤ TSN
– Low Latency, Predictability, and QoS
– TSN data flows naturally match DDS data flows
• DDS can leverage TSN to implement and guarantee Qos:
– Use TSN clock for time-stamping and global data ordering
– Automatically define TSN flows per DDS Topic
– Identify critical flows based on QoS (e.g. latency budget, deadline)
– Ensure QoS can be met (deadline, liveliness)
– Derive TSN schedule from DDS configuration & deployment

Use TSN as a deterministic transport for DDS over Ethernet

©2019 Real-Time Innovations, Inc.


Conclusion

©2019 Real-Time Innovations, Inc.


Edge Autonomy Requirements
• Scalability
– Can’t move and store everything in the cloud
• Real-Time / Real-World Performance
– Fast decision needed fast for robust autonomous systems
• Safety, Robustness & Availability
– Work continuously, resilient to infrastructure failures
• Security/Privacy
– Adapt to different security requirements while maintaining efficiency
• Entirely new applications based on AUTONOMY

©2019 Real-Time Innovations, Inc.


Edge Autonomy Presents a Huge Opportunity

©2019 Real-Time
©2019
Innovations,
Real-Time Inc.
Innovations,
Confidential.
Inc.
©2019 Real-Time Innovations, Inc.
Thank you!

©2019 Real-Time Innovations, Inc.


Stay Connected
rti.com rtisoftware
Free trial of Connext DDS

@rti_software connextpodcast

@rti_software rti.com/blog

©2019 Real-Time Innovations, Inc.

You might also like