Transcript OMG Smart Sensors Specification

HRTC Meeting
12 September 2002, Vienna
Smart Sensors
Thomas Losert
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.1
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.2
Smart Transducer (ST)
Comprises the Integration of one or more
Sensor/Actuator Elements with a Microcontroller and a Communication Network
Interface that provides the following
Services across standard Interfaces:




Diagnostic and Management
Real-Time Communication
Calibration of Sensor
Signal Conditioning and Conversion to
standard Units
Thomas Losert
12 Sept. 2002 / p.3
Advantages
 No noise pickup from long external signal
transmission lines.
 Better Diagnostics – Simple external ST failure
modes (e.g., fail-silent)
 "Plug-and-play" capability if the ST contains its own
documentation on silicon or in an external database.
 Reduction of the complexity at the system hardware
and software and the internal ST failure modes can
be hidden from the user by a well-designed fully
specified ST interface.
 Cost reduction in installation and maintenance.
Thomas Losert
12 Sept. 2002 / p.4
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.5
Observations
An observation consists of an atomic tuple
< observed value, instant, RT-entity Name >
For Communication of Observations across an RS
interface a common set of concepts is necessary:




Common representation of values in a Shared code-space
Common notion of time and its representation
Common meaning of the names of RT entities
Access protocol to the information.
Thomas Losert
12 Sept. 2002 / p.6
Observations
An observation consists of an atomic tuple
< observed value, instant, RT-entity Name >
For Communication of Observations across an RS
interface a common set of concepts is necessary:




Common representation of values in a Shared code-space
Common notion of time and its representation
Common meaning of the names of RT entities
Access protocol to the information.
Thomas Losert
12 Sept. 2002 / p.7
State vs. Event Values
Characteristic
State
Observation
Event
Observation
Value
Full Value
Value Difference
Frequency
Periodic
Sporadic
Loss of Observ. Period lost
Loss of synch.
Semantics
At-least-once
Exactly-once
Error Detection
At receiver
At sender only
Thomas Losert
12 Sept. 2002 / p.8
Observations
An observation consists of an atomic tuple
< observed value, instant, RT-entity Name >
For Communication of Observations across an RS
interface a common set of concepts is necessary:




Common representation of values in a Shared code-space
Common notion of time and its representation
Common meaning of the names of RT entities
Access protocol to the information.
Thomas Losert
12 Sept. 2002 / p.9
GPS - Time
GPS weeks and GPS seconds since the GPS epoch
(6. Jan. 1980, 00:00:00 UTC)
Advantages:
• Continuous representation of Time
(no leapseconds)
• High availability with low jitter
Disadvantages:
• Overflow every 19.6 years because GPS week is
represented as a 10 bit value
Thomas Losert
12 Sept. 2002 / p.10
Global Notion of Time
239 seconds
1 sec (20 seconds)
2-24 seconds
external time format (8 bytes)
Based on GPS; advantages of GPS are preserved
Additional advantages:
• Enhanced granularity (about 60 ns)
• Easy Access to the full second
• No overflow during the lifetime of the product
(horizon of more than 10 000 years)
• Easy manipulation of timestamps
Thomas Losert
12 Sept. 2002 / p.11
Observations
An observation consists of an atomic tuple
< observed value, instant, RT-entity Name >
For Communication of Observations across an RS
interface a common set of concepts is necessary:




Common representation of values in a Shared code-space
Common notion of time and its representation
Common meaning of the names of RT entities
Access protocol to the information.
Thomas Losert
12 Sept. 2002 / p.12
Interface File System (IFS)
Hierarchical, distributed File-System that is the
Source and Sink of each communication
Map all relevant properties of a node into the IFS:
• Sensor and actuator data
• Calibration data, parameters, etc.
• Configuration Information
• Node identification (physical name, global name)
• Documentation and pointer to register service
Thomas Losert
12 Sept. 2002 / p.13
Global Name
Is the concatenation of
Cluster Name,
Logical Name
File Name, and
Record Name
•
•
•
•
Up to 250 Clusters
Up to 250 Nodes
Up to 64 Files
Up to 256 Records
CORBA-Gateway
Active
Master
Active
Master
Active
Master
Cluster A
Cluster B
Cluster C
Thomas Losert
Transducer Node
12 Sept. 2002 / p.14
Observations
An observation consists of an atomic tuple
< observed value, instant, RT-entity Name >
For Communication of Observations across an RS
interface a common set of concepts is necessary:




Common representation of values in a Shared code-space
Common notion of time and its representation
Common meaning of the names of RT entities
Access protocol to the information (see TTP/A).
Thomas Losert
12 Sept. 2002 / p.15
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.16
Three Interfaces of a Node
Thomas Losert
12 Sept. 2002 / p.17
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.18
Why Time-Triggered?
• Requirement to record value and instant
when measurement was performed
• Requirement to perform synchronous
distributed measurements
• Requirement to gather multiple
measurements with bounded delay
Thomas Losert
12 Sept. 2002 / p.19
Principles of Operation: Temporal Firewall
•
•
•
•
•
•
Best way of communication for Producer
Best way of communication for Consumer
Predictable time-triggered communication
Components have access to a global time
Communication and measurement synchronized
Temporal firewall in communication
Thomas Losert
12 Sept. 2002 / p.20
Time-Triggered Protocol/Class A
(TTP/A)
• Low-cost time-triggered sensor bus
• Supports „Smart Transducers“
• Currently subject of a standardization
process at Object Management Group
Thomas Losert
12 Sept. 2002 / p.21
Principles of Operation: TTP/A (1)
•
•
•
•
One active master
Up to 250 slaves
Communication organized in rounds
TDMA bus allocation
Thomas Losert
12 Sept. 2002 / p.22
Principles of Operation: TTP/A (2)
• Time-Triggered Protocol for Fieldbus Applications
• TDMA Bus Access Scheme
• Communication Organized into Rounds
• Supports Various Physical Layers
Each round is initiated by the master by sending the
Fireworks-Byte
• Multipartner (MP): Real-Time Service Interface
• Master/Slave (MS): Monitoring and Configuration
Thomas Losert
12 Sept. 2002 / p.23
TTP/A: Principles of Operation (2)
Thomas Losert
12 Sept. 2002 / p.24
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.25
Architecture
Thomas Losert
12 Sept. 2002 / p.26
CORBA
CORBA Smart Transducers Interface (STI)
• Allows (re)configuration of the Smart
Transducer cluster, e.g. over the internet
• Allows monitoring of the ST cluster
• Allows access to the real-time service
interface for cases, when a RT-CORBA
supporting end-to-end predictability is
used
Thomas Losert
12 Sept. 2002 / p.27
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.28
Definition of Sensor Fusion
„Sensor Fusion is the combination of sensory
data or data derived from sensory data such
that the resulting information is in some
sense better than would be possible when
these sources were used individually.“
Thomas Losert
12 Sept. 2002 / p.29
Purposes of Sensor Fusion
„better“ can mean:
• More robust (fault-tolerant sensor clusters)
• Extended spatial and temporal coverage
• Increased confidence
• Reduced ambiguity and uncertainty
• Robustness against interference
• Improved resolution
• Emerging services
Thomas Losert
12 Sept. 2002 / p.30
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.31
A Typical Sensor Fusion Control Loop
Thomas Losert
12 Sept. 2002 / p.32
Introducing a Sensor Fusion Model
Three levels of computation:
• Node level
• Cluster level
• Control application level
Well-defined interfaces between levels
• Smart Transducer Interface
• Environmental image
Thomas Losert
12 Sept. 2002 / p.33
Control Loop in the Three-Level Model
Thomas Losert
12 Sept. 2002 / p.34
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.35
Case Study:
Autonomous Mobile Robot
• TTP/A fieldbus network with 1 master
• 1 node for sensor fusion and navigation
• 15 smart transducers (9 actuators/6 sensors)
Thomas Losert
12 Sept. 2002 / p.36
ST-Types
• 2 Ultrasonic Far Distance Sensors
• 3 Infrared Low Distance Sensors
• 1 Luminance Sensor
•
•
•
•
1 Steering Control
1 Speed Control
3 Servos for Moving Infrared Sensors
4 Lighting Control
Thomas Losert
12 Sept. 2002 / p.37
Control Loop
Thomas Losert
12 Sept. 2002 / p.38
The „Smart Car“
Thomas Losert
12 Sept. 2002 / p.39
Overview
• Smart Transducers
(Definition, Observations, Interfaces)
• Principles of Operation
• Architecture
• Sensor Fusion (Definition, Concepts)
• Case Study
• Conclusion & Outlook
Thomas Losert
12 Sept. 2002 / p.40
Conclusion & Outlook
Case Study
• Proof of concept for TTP/A
• Testbed for Sensor Fusion
• Demonstrator for CORBA-STI
The Standard Proposal has been adopted in
Jan. 2002 by the Board of Directors and is
called Smart Transducers Interface Standard
now. The standardization process continues ...
Thomas Losert
12 Sept. 2002 / p.41
Thank you for your attention!
Thomas Losert
12 Sept. 2002 / p.42