Multi-Agents Supporting Reflection in a Middleware for Mission-Driven Heterogeneous Sensor Networks...

Post on 22-Dec-2015

218 views 0 download

Tags:

Transcript of Multi-Agents Supporting Reflection in a Middleware for Mission-Driven Heterogeneous Sensor Networks...

Multi-Agents Supporting Reflection in a Middleware for Mission-Driven Heterogeneous Sensor Networks

Edison Pignaton de FreitasMarco Aurélio Wehrmeister

Armando Morado Ferreira

Carlos Eduardo Pereira

Tony Larsson

3rd ATSN @ 8th AAMAS – May 2009

Outline

• Context• Proposed Approach • Mission-driven Approach

• Mission Description Language• Mission Parameterization

• Middleware Overview• Planning-agent Model• In-network Reasoning• Conclusion and Future Work

Context

• Emerging Sophisticated Sensor Networks Applications• Heterogeneous sensors, great number of nodes, the

need for distributed decisions, dynamic scenarios …

• Dynamic scenarios• Environment conditions, topology changes, …

• System Life Time• Requirements may change

Context

• Motivation Application Scenarios

Surveillance and Patrolling

Rescue Assistance and Disaster Recovering

Context

• Problems to address:• Establishment of a network mission (and its partitioning)• Efficient use of different types of sensors in the network• Data aggregation/fusion • Management of nodes, groups, clusters and the whole

network • Task (Re)Allocation

• Dynamically changes (reachability, capabilities, remaining resources, etc)

Our Focus

This presentation

Proposed Approach• System Overview

• High-level Mission Description Language• Middleware providing interoperability• Agent-based support:

• Mission dissemination and network reasoning

Mission-driven Approach

• Mission Description Language (MDL): high-level mission statements • Text-based, maps and other representations… • Simple Example:

IF DETECT <DECREASE_OF <temperature>> WITH GRANULARITY<3> MONITOR <FOG>WITH ACQUISITION <period = yy>

• MDL Translation:

Mission-driven Approach

Mission is carried by mobile agents…

• Mission Parameterization: Formal representation

QFMMSNSMGM ,,,tempi < tempi-1 – 3

tempSensor <?s?>humSensor <?s?>

Qf(mi,s) = q

Mission-driven Approach

SM = set of node-missions mi = node-mission (set of measurements SME + a set of constraints SMC)SN = set of sensorsMM = mission-mappingQF = quality function

m1 =fog(t,h), per = yy

m2 =

SN =

MM = mm(mi) = s

QF =

SM = SME = temp, humSMC = per, threshold

Middleware Overview

• Service division• Types of agents

• Planning-: missions management

• Mission-: mission carrier

• Service-: service provider

• Adaptation • Reflection• Mission execution

Planning-agent Model

• BDI model• Beliefs: background info, node status, and

environment status• Ex: the current temperature, previous temperature, QF

value for a given node-mission

• Desires: goals related to the node-missions assumed

• Ex: trigger the fog sampling when tempi < tempi-1 -3

Planning-agent Model• BDI model

• Intentions: what have to be done to accomplish with the above goals (SME), respecting the constraints (SMC)

• Ex: measure temperature and compare with the previous value, according the threshold, respecting the sampling rate…

• Plan: sequence of actions to accomplish the goals according its intentions

• Ex: 1) Acquire temperature sample from the device; 2) Store sample, 3) Compare current sample with stored value; 4) if rule, send message to HumSensor …

Planning-agent Model

• Architectural Structure

In-network Reasoning

• Autonomous negotiation mechanism to distribute the node-missions among nodes – Mission Setup

• Evaluation mechanism to assess the efficiency of the mission accomplishment and changes that must take place – Mission Adaptation

In-network Reasoning

• Mission Setup• Local decision about node-mission distribution• Context-awareness• 4-step simple mechanism

• 1st: mi’s SME and SMC analysis (partial belief)

• 2nd: nodes candidacy • 3rd: best effort candidacy• 4th: others candidacy analysis using QF (common belief)

In-network Reasoning

• Mission Adaptation• Node conditions and environmental changes

awareness (updates in nodes’ beliefs)• Two cases considered:

• Node failure: as soon as perceived by the neighbor nodes

• Node is not able to continue the node-mission or another can perform it better: QF based decision

In-network Reasoning

• Considerations about Complexity• Mechanisms are customized to fit the resource

budget of the different types of nodes• Simpler nodes have simpler QF• The internal parts of the planning-agent

architecture are also customized for each kind of node, considering more or less parameters according the nodes’ capabilities and resources

Future Work

• Current in fact… • Simulations using ShoX

• Java based• Network concepts (OSI, signal propagation, interfs… )• Mobility models• Energy Cons/Prod models

Future Work

• On going • Simulations using ShoX

• Evaluation of the cost of the proposed approach• Introduction of the agents’ model in the tool framework

• Future• Interface with the Mission Specification Console

and ShoX• Complete simulation from the MDL to the runtime

agents’ (re)negotiation

Conclusion

• Presentation of a methodology to address heterogeneous sensor networks from high-level directions to autonomous nodes decisions

• A high-level language translated in system parameters

• Middleware and agents to support the proposed mission-driven approach

• A BDI-based agent model to provide the required reasoning features

Thanks for your attention!

Questions ? Suggestions?

Supervisors: tony.larsson@hh.se cpereira@ece.ufrgs.br

PhD Student: edison.pignaton@hh.se

Backup slides

• Heterogeneity• Mission Dissemination• In-network reasoning

HeterogeneityComputerPlatform

Mobility

Sensor Capabilities

WSN

HPS

HPS: High-performance SensorsMANET: Mobile Adhoc NetworkVANET: Vehicle Adhoc NetworkWSN: Wireless Sensor Network

MANET HeterogeneityCube

2D

3D

StaticLow High

High

VANET

Mission Dissemination

• Mobile agents: Mission-agents

After the mission translationMobile-agents disseminate itin the network

Area described in the mission

MissionAgentsPlanning Agent

Mission Dissemination

• Mobile agents: Mission-agents

One time in the network, they follow the path to the area defined in the mission directions, by moving and/or cloning. (1)

Mission Dissemination

• Mobile agents: Mission-agents

One time in the network, they follow the path to the area defined in the mission directions, by moving and/or cloning. (2)

move

move

Mission Dissemination

• Mobile agents: Mission-agents

One time in the network, they follow the path to the area defined in the mission directions, by moving and/or cloning. (3)

move

clone

Mission Dissemination

• Mobile agents: Mission-agents

One time in the network, they follow the path to the area defined in the mission directions, by moving and/or cloning. (3a)

clone

Now the mission is disseminated!

In-Network Reasoning

• Multi-agents reasoning: mission setup

According to mission requirements and nodes capabilities, the node-missions are divided among the nodes needed to accomplish the mission.

Negotiation

NegotiationN

egotiation

In-Network Reasoning

After the negotiation nodes have divided the job that has to be done to accomplish the mission and nodes that are not needed to be employed in the mission, dealocate the respective mission-agent.

Dealocate

WorkDivided!

• Multi-agents reasoning: mission setup

• Multi-agent reasoning: adaptations

Changes in the environment require re-negotiation to decide which node will execute each task:- Node’s capabilities- Actual state- Task requirements

Renegotiation

In-Network Reasoning

In-Network Reasoning

• The negotiation is carried out by means of a light-weight protocol in order to not overload the network with control messages (4-step protocol presented)

• Decision making is based on a quality function, that measures how good a node can perform a given mission (or node-mission). In fact it measures the utility in use a node, or a set of them, in order to perform the tasks needed to accomplish a given mission (or node-mission). Ex.:

)))(),(()),(),(),((),((max)( ,,max tMtESEvtptMtfCtUtU jiii

jii

Skji

TSKi

TSK jj Utility function for the UAVs based on: the employability of the sensor device, the proximity of the node to the phenomenon, environmental influences (e.g. weather conditions) and remaining resources.

Adaptation

• Mobile-agents: Service-agents

Mobile-agents providing Services at different places:- Move/Clone

X

Adaptation

• Mobile-agents: Service-agents

Mobile-agents providing Services at different places:- Move/Clone

Clone

Move