Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
.
.
. ..
.
.
Towards Discrete Control for the Internet of Thingsand Smart Environments
Mengxuan Zhao, Gilles Privat, Orange Labs, Grenoble, FranceEric Rutten, INRIA, Grenoble, France
Hassane Alla GIPSA Lab, Grenoble, France
June 25, 2013
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 Background
.. .3 IoT and smart environments
.. .4 Modelling as a DCS problem
.. .5 Simulation
.. .6 Conclusion
2/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 Background
.. .3 IoT and smart environments
.. .4 Modelling as a DCS problem
.. .5 Simulation
.. .6 Conclusion
3/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Motivation.IoT and smart environments..
.
. ..
.
.
massive instrumentation, networked sensors and actuatorsoutside-in robots, sense and act upon their own inner spacesmart homes, smart buildings or smart cities
.Control techniques for autonomic management..
.
. ..
.
.
model possible behaviours, and control objectives, separatelyclassically continuous time dynamics and differential equationsdiscrete control, events and states, Petri nets or automata
.Discrete control for the IoT..
.
. ..
.
.
systematic modelling framework ; case study in smart homeautomata ; Discrete Controller Synthesis (DCS)
4/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 BackgroundInterfacing to the IoT and Smart EnvironmentsReactive languages, DCSDiscrete control as MAPE-K
.. .3 IoT and smart environments
.. .4 Modelling as a DCS problem
.. .5 Simulation
.. .6 Conclusion
5/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Interfacing to the IoT and Smart Environments
Entity Entity Entity Entity
SupervisoryController 1
SupervisoryController 2
Application 1 Application 2
Sensor/Actuator
Sensor/Actuator
Sensor/Actuator
Other Service
Space Space Thing Thing
ApplicationLayer
ServiceLayer
Entity AbstractionLayer
DeviceLayer
PgysicalPlant
.monitoring and controlling of entities (rooms, appliances, ...)..
.
. ..
.
.
intermediate abstraction layer HAL (Home Abstraction Layer)supervisory controllers as service : emphasis on genericity
6/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Reactive languages, synchronous programming.Modelling formalism and programming language..
.
. ..
.
.
reaction to input flows → output flowsdata-flow nodes and equations ; mode automata (FSM)parallel (synchronous) and hierarchical composition
synchronous languages, (25+ years)tools: compilers (e.g., Heptagon), code generation, verification, ...
example: delayable task control (in Heptagon)delayable(r,c,e) = a,s
Idle Wait
e r and c/s
Activec/s
r and not c
a = true
a = falsea = false node delayable(r,c,e:bool) returns (a,s:bool)
let automaton
state Idle do
a = false; s = r and c
until r and c then Active
| r and not c then Wait
state Wait do a = false; s = c
until c then Active
state Active do a = true; s=false
until e then Idle
end tel
7/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Discrete controller synthesis (DCS): principle
.Goal..
.
. ..
.
.
Enforcing a temporal property Φ on a systemon which Φ does not yet hold a priori
.Principle (on implicit equational representation)..
.
. ..
.
.
State memoryTrans transition functionOut output function
Partition of variables : controllable (Y c), uncontrollable (Y u)Computation of a controller such that the controlled systemsatisfies Φ (invariance, reachability, attractivity, ...)
DCS tool: Sigali (H. Marchand e.a.)
8/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Discrete controller synthesis (DCS): principle
.Goal..
.
. ..
.
.
Enforcing a temporal property Φ on a systemon which Φ does not yet hold a priori
.Principle (on implicit equational representation)..
.
. ..
.
.
State memoryTrans transition functionOut output function
Trans State OutZY
Partition of variables : controllable (Y c), uncontrollable (Y u)Computation of a controller such that the controlled systemsatisfies Φ (invariance, reachability, attractivity, ...)
DCS tool: Sigali (H. Marchand e.a.)
8/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Discrete controller synthesis (DCS): principle
.Goal..
.
. ..
.
.
Enforcing a temporal property Φ on a systemon which Φ does not yet hold a priori
.Principle (on implicit equational representation)..
.
. ..
.
.
State memoryTrans transition functionOut output function
Y c
Y u Trans State OutZY
Partition of variables : controllable (Y c), uncontrollable (Y u)
Computation of a controller such that the controlled systemsatisfies Φ (invariance, reachability, attractivity, ...)
DCS tool: Sigali (H. Marchand e.a.)
8/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Discrete controller synthesis (DCS): principle
.Goal..
.
. ..
.
.
Enforcing a temporal property Φ on a systemon which Φ does not yet hold a priori
.Principle (on implicit equational representation)..
.
. ..
.
.
State memoryTrans transition functionOut output function
Ctrlr Y c
Y u Trans State OutZY
Partition of variables : controllable (Y c), uncontrollable (Y u)Computation of a controller such that the controlled systemsatisfies Φ (invariance, reachability, attractivity, ...)
DCS tool: Sigali (H. Marchand e.a.)
8/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
BZR programming language [http://bzr.inria.fr]
.
.
. ..
.
.
built on top of nodes in Heptagonto each contract, associate controllable variables, localat compile-time (user-friendly DCS),
compute a controller for each componentwhen no controllable inputs : verification by model-checkingstep and reset functions ; executable code : C, Java, ...
OutCTrC StC
Trans StateCtrlr
eA, eG
xi
contract
Outyj
ck
body
twotasks(r1, e1, r2, e2) = a1, s1, a2, s2enforce not (a1 and a2)with c1, c2
(a1, s1) = delayable(r1, c1, e1);
(a2, s2) = delayable(r2, c2, e2)
& G. Delaval & H. Marchand [ACM LCTES’10] [jDEDS13]9/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Discrete control as MAPE-K
sensor
executeknowledge
monitor
analyse plan
actuator
managed element
.
.
. ..
..
autonomic MAPE-Kflows : sensor observationsto reconfiguration actionsreactive language BZR usedas DSL for decision
sensor
stateinputs
actuator
managed element
outputs
transitionfunction
control
.
.
. ..
.
.
FSM instanciation ofMAPE-Kexhibit state (observability)accept events or conditions(controllability)
10/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Hierarchical architecturecontrol
stateinputs outputs
transitionfunction
sensor
stateinputs
actuator
outputs
transitionfunction
sensor
stateinputs
actuator
outputs
transitionfunction
controlcontrol
managed element
.
.
. ..
.
.
hierarchical MAPE-K, through additional interfacesin components : composites using life-cycles of subcomponentsimplementation : step
synthesized and generated off-linecalled at run-time in composite controller
11/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 Background
.. .3 IoT and smart environmentsTarget environmentsReconfiguration policy
.. .4 Modelling as a DCS problem
.. .5 Simulation
.. .6 Conclusion
12/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Target environments
lamp
room
window
radiator
door
TV washingmachine
oven
sensor readings
actuator commands
Controller
SystemModel
abstractedby HAL
state
cm
cn
ck
.Smart home example..
.
. ..
.
.
set of entities, observation and control possibilitiesissues of safety (source of light on in case of presence),issues of economy or comfort (instantaneous power peaks)
13/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Reconfigurable entities
.Basic entities in the apartment..
.
. ..
.
.
door, windowlamp, TV, radiator (off, frost-free, eco or high mode)oven : off, heating up, maintaining its current temperaturewashing machine : phases ; can be suspended
.Sensors and actuators..
.
. ..
.
.
presence, door/window (open or closed)smart plugs for TV or lamp, suspend the wash machine
14/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Reconfiguration policy.4 categories of objectives..
.
. ..
.
.
safety (sa)security (se)energy efficiency (e)comfort (c)
.System objectives..
.
. ..
.
.
...1 (sa) at least one light source on when room is occupied
...2 (se) close window and door when room isn’t occupied
...3 (e) if window or door open, radiator either off or frost-free
...4 (e) if room inoccupied, no light on and radiator off or frost-free
...5 (sa,e) total power under current threshold(3 modes: minimal-safety, comfort, eco)
15/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 Background
.. .3 IoT and smart environments
.. .4 Modelling as a DCS problemSystem behaviors modelControl objective specification
.. .5 Simulation
.. .6 Conclusion
16/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
System behaviors model (i)
closed openpushi & c_doori
pushi or not c_doori door_openi
= truedoor_openi = false
pushi
c_doori door_openi
doori
empty ocupied
presencei
not presencei room_oci = true
room_oci = false
roomi
presencei room_oci
.door and room..
.
. ..
..
two states /configurationscontrollability
door : preventopening, closeroom : observeronly
offfrost
protection
up1
c1
state
rad
down1 eco
power
(0) (300) (1500)
high(2000)
up2
down2
up1 or not c1
down1 or not c1
up1 & c1
down1 or not c1
down2 or not c2
up2 & c2
up2 & c2
down1 ornot c1
c2
.radiator / heater..
.
. ..
.
.
four states / configurationscontrollability
prevent going higherforce going lower
17/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
System behaviors model (ii)
offwater
fill
washing
rinse
spin
standby
startend& not c
end& c
end(250)
(200)
(800)
(0)
washing machine
end
start
c
state
power
end
(0)
end
(15)
c
.washing machine..
.
. ..
.
.
phases, with powerstandby controllable
off
heat
maintain reheat
standby
start& not c
start& c
temp_ok
(1500)
(300)
(0)
oven
finish
start
c state
powercold (0)
(1000)
temp_ok
finish
finish finish
c
cold & c
temp_ok
.oven..
.
. ..
.
.
5 states, with powerstandby, reheatcontrollable
18/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
System behaviors model (iii)
minimalsafety
eco_inputeco
management policy
comfort_input
eco
comfort
comfort
eco_input
not eco_input
eco_inputnot
comfort_input
comfort_input
comfort_input
PL=5000
PL=7000
PL=3000
PL
.Global system behavior model..
.
. ..
.
.
parallel composition of instances of automataglobal power consumption functions, in terms of the local onestotalPower = p(wm) + p(ov) + p(rd)
3 management policies : 3 states with different PL
19/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Control objective specification
.Conjunction of five rules..
.
. ..
.
.
...1 room_oc ⇒ lamp_on ∨ tv_on
...2 ¬room_oc ⇒ ¬(d_open ∨ w_open)
...3 (d_open ∨ w_open) ⇒ (rad_off ∨ rad_frost)
...4 ¬room_oc ⇒ (¬(lamp_on∨tv_on)∧(rad_off ∨ rad_frost))
...5 totalPower ≤ PL
made invariant by control ; controller synthesizedtool-supported : BZR
20/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 Background
.. .3 IoT and smart environments
.. .4 Modelling as a DCS problem
.. .5 Simulation
.. .6 Conclusion
21/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Implementation and simulation
.BZR encoding and DCS..
.
. ..
.
.
textual syntax for automata and contractsexecutable code generation (C or Java)
.Simulation..
.
. ..
.
.
MiLeSEnS (Multi-Level Smart Environment Simulator)sensor and actuator models
22/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Outline
.. .1 Motivation
.. .2 Background
.. .3 IoT and smart environments
.. .4 Modelling as a DCS problem
.. .5 Simulation
.. .6 Conclusion
23/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Conclusions
.DCS for IoT..
.
. ..
.
.
first results in using discrete supervisory controllersapplied to the IoT and smart homes and buildings
.Systematic modeling framework..
.
. ..
.
.
behavioral modeling of typical entitiesformulation of control objectives of typical categories
(safety, security, energy, comfort)automatic generation of controllersdevelopment and experimental validation
24/25
Motivation. . . . . .Background
. . .IoT and smart environments
. . . .Modelling as a DCS problem Simulation Conclusion
Perspectives.Perspectives..
.
. ..
.
.
Domain-Specific Language for smart environmentsgeneration of LTS and objectives
more DCS : modular, quantitative ; identification of modelsmore complete experiments
.Part of a general approach using discrete control for computing..
.
. ..
.
.
behavioral modeling using LTSs : high level
automatic generation (DCS) : safe, max.permissive
tool support (BZR language)used also for :
coordinating administration loops in the Cloud or HPCmanaging reconfigurable architectures (FPGA)
25/25