P. Albertos* & A. Crespo+

Universidad Politécnica de Valencia* Dept. of Systems Engineering and Control,

+ Dept. of Computer Engineering

POB. 22012E-46071 Valencia, Spain. Fax: +34 96 3879579

e-mail: pedro@aii.upv.es

Embedded Control Systems:Control Kernel

NoE on Embedded Systems Design

©P. Albertos 2005

Embedded Control Systems

• Embedded systems with:– hard RT constraints– guarantee of safe operation– best possible performances

• Additional issues from viewpoint of:– implementation– computation– algorithmic

NoE on Embedded Systems Design

©P. Albertos 2005

RT Control Issues • RT Constraints:

– Maximize the time determinism• For many controllers a worst-case approach works well e.g.,

PI, PID, State Feedback, …

however, many exceptions:• hybrid controllers that switch between different modes with

different characteristics• model-predictive controllers (MPC)• convex optimization problem solved every sample

execution time can vary an order of magnitude

– Compensate the variations:• Measure and react• Feedback robustness

NoE on Embedded Systems Design

©P. Albertos 2005

Control requirements

• Multiloop control• Non-uniform sampling• Missing data• Variable delays• Sampling period changes• Mode changes• Fault tolerant• Safe operation• CPU optimization• Battery control

Process to control

Environment

ECS

CPUBattery

S1

Si

Memory

A1

Ai

NoE on Embedded Systems Design

©P. Albertos 2005

ECS: Implementation• The same resources must be shared between

different tasks

• Alternative control algorithms should be ready to get the control of the process

• Working conditions, such as priority, allocated time and memory or signals availability may change

• Variable delays should be considered

• Validation and certification

NoE on Embedded Systems Design

©P. Albertos 2005

ECS: Computational viewpoint• Economic algorithms• Information updating • Optional tasks • Hybrid systems • CPU use measurement and optimisation• On-line scheduling• Memory saving • Economic hardware redundancy• Fault detection and isolation

NoE on Embedded Systems Design

©P. Albertos 2005

Control Performances

• RT Constraints:– Maximize the time determinism

• For many controllers a worst-case approach works well e.g., PI, PID, …

however, a lot of exceptions:• hybrid controllers that switch between different modes with

different characteristics• model-predictive controllers (MPC)• convex optimization problem solved every sample• execution time can vary an order of magnitude

– Compensate the variations:• Feedback robustness• Measure and counteract

• Relevance of the control actions

The Control Effort concept

• Sensitive to time delays

• Changes in the sampling period:– Controller parameters– Past data

• Commutation bumping

NoE on Embedded Systems Design

©P. Albertos 2005

MIMO controlled plant

ref

Hj.... .... ........

yiuj

vj

zi

Si

Process

Control

IntegrityRedundancyPerformance degrading

NoE on Embedded Systems Design

©P. Albertos 2005

ECS: Control algorithm viewpoint– Reduced order models – Non-conventional sampling and updating patterns– Missing data control– Event-triggered control– Hybrid control systems – Decision and supervisory control– Multimode control– Sampling rate changes– Fault-tolerant control – Degraded and back-up (safe) control strategies– Battery monitoring and control

NoE on Embedded Systems Design

©P. Albertos 2005

The kernel concept

• Basic services:– Task and time management

– Interrupt handling

– Interface to the applications (API)

– Mode changes

– Fault tolerance

OS kernel:

• Additional services– File management– Quality of service– Tracing and debugging

– Mode changes

– Fault tolerance

NoE on Embedded Systems Design

©P. Albertos 2005

The OS kernel (I)The OS Kernel provides the minimal services that should be included in any embedded system.

• Task management and synchronization mechanisms • Task communication• Semaphores and monitors• Server definition (aperiodic servers, constant bandwidth servers))• Scheduling policies

• Time management• Real-time clock• High resolution timers and execution timers (limit the cpu use per task)• Absolute and relative delays

• Application interface (API)• POSIX (standard) • OSEK (automobile industry specification)• ….

NoE on Embedded Systems Design

©P. Albertos 2005

The OS kernel (II)The OS Kernel provides the minimal services that should be included in any embedded control system.

• Fault tolerance • Degrade task activity (when a task does not guarantee some timing constraints, the degraded behavior is executed)• Change mode events raised when some faults can not be managed.

• Mode changes • Mode definition (set of tasks associated to a mode) • Mode change events (event to change from one mode to another) • Mode change protocol

• Security• Network access• Attacks

NoE on Embedded Systems Design

©P. Albertos 2005

Hardware

Interrupt services

Task management

Mode tasks

API

Application Tasks

The OS kernel (III)

NoE on Embedded Systems Design

©P. Albertos 2005

The Control Kernel concept

• Ensures control action (CA) delivering• Data acquisition of major signals• Transfer to new control structure

• Additional CA computing facilities• Communication facilities• Coordination facilities

DE driven-control

NoE on Embedded Systems Design

©P. Albertos 2005

Event-driven control

F

NoE on Embedded Systems Design

©P. Albertos 2005

Basic Control Loop

A/D

ControlAlgorithm

D/A

referencer(t)

Sensor ActuatorProcess

y(t) u(t)

ukyk

rkA/D

Regulator

CT

DT

NoE on Embedded Systems Design

©P. Albertos 2005

Basic Control Loop

task body Controlador is Periodo : Time_Span; -- duracion Proxima_Iteracion : Time; -- tiempo absoluto begin Proxima_Iteracion := Clock; loop convertir_sensor_analogico_digital(y); calcular_accion_control(u); enviar_accion_control_convertida(u); actualizar_variables_internas(e,y,u,…); Proxima_Iteracion:=Proxima_Iteracion + Periodo; delay until Proxima_Iteracion; end loop; end Controlador;

NoE on Embedded Systems Design

©P. Albertos 2005

Control Kernel• Ensures control action (CA) delivering

– Safe (back-up) CA computation– Safe CA computation based on previous data

• Data acquisition of major signals– Safe CA computation based on current data

• Transfer to new control structure– Basic control structure parameters computation– CA computation

• Full DA – Control structures evaluation and selection– CA computation (different levels)

• Communication facilities• Coordination facilities

NoE on Embedded Systems Design

©P. Albertos 2005

The control kernel concept

kvku

Plant

CAdeliv

Backup CA CAuser

CAcomputation

bu

kx

kmky

DA

kr

Safe operation in any condition

NoE on Embedded Systems Design

©P. Albertos 2005

Control Kernel Algorithm

• CA delivering • Backup CA bk uu

• Backup CA Computation ),( 1 kbk xufu

• Current safe b-up CA comp.

kk uv

• Basic CA computation ),(1 kkk xrfu

),( kbk xufu

• CA comp ),( kkik xrfu • CA comp. (Process model)

– Essential– Partial– Complete

),( kkk xrFu

NoE on Embedded Systems Design

©P. Albertos 2005

Control Kernel Algorithm

• Model reduction:– Partial control (parts

of the plant)

modes slow : modes;fast : 21 xx

- Partial phenomena (fast/slow dynamics)

S1

S2

u y

k

kk

kkx

xCCu

B

B

x

x

AA

AA

x

x

2

121

2

1

2

1

2221

1211

12

1 y ;

NoE on Embedded Systems Design

©P. Albertos 2005

Conclusions

• Control Kernel

• Variables relevance– Control effort