Click here to load reader
Upload
jose-luis-poza-lujan
View
466
Download
0
Embed Size (px)
DESCRIPTION
Citation preview
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Event Management Proposal for Distribution Data Service StandardJosé-Luis Poza-LujánJuan-Luis Posadas-YagüeJosé-Enrique Simó-TenUniversity Institute of Control Systems and Industrial Computing (ai2)Universitat Politècnica de València (UPV).
···
Events
···
Actions
···
···
···
···
···
···
···
Conditions
Component Operations
QoS Alarms
QoC Alarms
Message Filters
Component Functions
QoS Configuration
QoC Configuration
Message Filtering
Event
-m_iState-m_iType-m_pCondition
+setState()+getState()+getType()
EventAlarm
-m_iAlarmType-m_iAlarmValue-m_iStartupAlarm-m_iThersholdRising-m_iThersholdFalling
EventFilter
-m_iOffset-m_sPattern-m_sMask-m_bSendEvent-m_bDeleteMessage-m_bCloseQueue
Entity
-m_pListOfConditions
+onCondition()+addCondition()+removeCondition()
Condition
-m_iState-m_pListOfEvents-m_pEntity-m_iPriority
+setState()+getState()+onEvent()
EventComponent
-m_iEventComponentType
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
NCS, EBC and Middleware• Event-Based Control (EBC) paradigm
(also called event-driven control) is adopted to implement systems where…– The periodic sampling is not possible.– The periodic sampling is not recommended
• In networked control systems (NCS), distributed control elements are connected by a network.– Control elements needs to know the
technology– Problems with synchronization
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
NCS, EBC and Middleware• A middleware enhances and offers to
control elements a set of services in order to facilitate the access to the network. – If EBC is used in NCS, the middleware will be
an essential component.• Middleware architecture is based on
communications paradigms.– Message passing– Client-server– Publish/Subscribe (P/S)– Blackboard.
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
DDS Model• Data Distribution Service for Real-Time
Systems (DDS) is an OMG standard middleware based on the P/S paradigm.– Offers time-controlled communications
between components using Quality of Service (QoS) polices.
Do
mai
n P
arti
cip
ant
write read, takeon_data_available
Producer Consumer
Data Writer
Data Writer
Publ
ishe
r
Data Reader
Listener
Subs
crib
er
Topic
Data Writer
Data Reader
Subs
crib
er
Publ
ishe
r
Data Reader
Topic
Topic
Do
mai
n P
arti
cip
ant
Do
mai
n P
arti
cip
ant
read, take write
ConsumerProducer & Consumer
Consumer
read, take
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
DDS Model• DDS can be used to send any kind of
Data, including event data. • DDS allows the application (control
component in NCS) to perform flexible filtering of events.
Entity
StatusCondition
1
0..1statuscondition
ConditionWaitSet* *
GuardCondition ReadCondition
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
DDS Model• DDS can be used to send any kind of
Data, including event data. • DDS allows the application (control
component in NCS) to perform flexible filtering of events.
• NCS needs certain event management,
But DDS does not define a built-in event type and advanced event management.
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Event management proposal• DDS offers adequate support to
coordinate communications between nodes, but it does not provide a mechanism to review the internal characteristics.
• To extend the capacity of DDS, a new component, called Action, is added.
• The role of Condition and Waitset components has been increased.
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Event management proposal• Main components are “Events”,
“Conditions” and “Actions”
···
Events
···
Actions
···
···
···
···
···
···
···
Conditions
Component Operations
QoS Alarms
QoC Alarms
Message Filters
Component Functions
QoS Configuration
QoC Configuration
Message Filtering
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Event management proposal• In our proposal Condition is similar to
WaitSet DDS element, Event is similar to Condition and Action is the new element
Entity
1
*condList
Event
EventAlarm EventFilterEventComponent
ConditioneventList
*
Action1 *
*actList
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Implementation• A distributed mobile robot navigation
environment has been used to test the validity and usefulness of the proposed model.
• The mobile robot is controlled by a set of control algorithms.– Reactive behaviours algorithms, are embedded
within the robot.– The deliberative algorithms, are implemented in
distributed nodes• In the experiment, all algorithms are
implemented in distributed nodes
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Implementation• The algorithm used is the “obstacle
avoidance” based on Braitenberg vehicles behaviours.
Mleft Mright
S0
S1
S2 S3
S4
S5
S6S7
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Implementation• Obstacle avoidance algorithm
1. Sensor detects obstacle and sends the distance to control node with a specified period
2. Control node changes the robot speed and path in function of the distance detected by all sensors in the ring using the formula to each motor
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Implementation• Optimization with event management
– The DDS based middleware filters sensor messages and only sends significant messages. The time interval between messages can change the robot speed by means the formula
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Experiment and results• Two scenarios have been tested
a) Robot in corridorb) Wall in front of robot
• Event Management influence1) Robot navigation without middleware event
detection 2) The same navigation with event detection and one
Action linked by means a Condition
Scenarios(1) DDS (2) DDS Event
ManagementSamplingPeriod
DistanceAverage
SamplingPeriod
DistanceAverage
(a) Robot in corridor 10.0 2.1 62.6 2.1
(b) Wall in front of robot 10.0 1.1 9.4 0.6
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
a) Robot in corridor– No significant
differences in distance– Message load
optimization
b) Wall in front of robot– Significant differences
in distance (speed optimization)
– Message load optimization
Experiment and results
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Contextualization
DDS Model
Event management proposal
Implementation
Experiment and results
Conclusions and future work
Conclusions and future work• Proposal to increase the event management
system proposed in the DDS standard• The most significant contribution is the
inclusion of a new object called Action.– Actions automatically make changes on the
middleware based on a combination of events.• The Action object has been tested with a
simple mobile robot system. • Future work
– Test the middleware with complex event combinations to generate different actions.
– Use priorities in Conditions or in Actions.
10th International Symposium on Distributed Computing and Artificial IntelligenceSalamanca, 22nd-24th May, 2013
Thanks for your attention
···
Events
···
Actions
···
···
···
···
···
···
···
Conditions
Component Operations
QoS Alarms
QoC Alarms
Message Filters
Component Functions
QoS Configuration
QoC Configuration
Message Filtering
Event
-m_iState-m_iType-m_pCondition
+setState()+getState()+getType()
EventAlarm
-m_iAlarmType-m_iAlarmValue-m_iStartupAlarm-m_iThersholdRising-m_iThersholdFalling
EventFilter
-m_iOffset-m_sPattern-m_sMask-m_bSendEvent-m_bDeleteMessage-m_bCloseQueue
Entity
-m_pListOfConditions
+onCondition()+addCondition()+removeCondition()
Condition
-m_iState-m_pListOfEvents-m_pEntity-m_iPriority
+setState()+getState()+onEvent()
EventComponent
-m_iEventComponentType
The study described in this paper is a part of the coordinated project COBAMI: Mission-based Hierarchical Control. Education and Science Department, Spanish Government. CICYT: MICINN: DPI2011-28507-C02-01/02