Embedded Real-Time Systems. Facts and figures. Characteristics

Transcription

1 Embedded Real-Time Systems Properties of embedded real-time systems Types of functionality Many different types coexist Run-time support Real-time Operating systems (RTOS) Modelling embedded real-time systems Task models Analsysis methods Distributed embedded systems Facts and figures 98% of all manufactured CPUs are used in embedded systems 70% of development cost for embedded complex systems are due to software development [A. Sangiovanni- Vincentelli, 1999] Industrial embedded systems are complex Automotive vehicles 40 processors Robotics 2.5 million lines of code 10 km print-outs Characteristics Vital quality attributes Reliability Availability Safety Openess integration of third party functionality Predictable timing behaviour

2 Functionality in embedded systems Types of functionality in a vehicle Feedback control systems - engine control, Anti-slip, ABS brakes, and vehicle dynamics. Discrete control systems Lights, window control, locks. Diagnostic systems Infotainment systems - TV, video, gaming, navigation systems Telematic systems - automatic emergency calls, Internet access In a not so distant future Safety critical control systems - X-by-wire systems, brakeby-wire, steer-by-wire. Computer controlled process The engineering goal is to control some aspect of the process This have traditionally been done by mechanical and electro mechanical solutions To reduce cost and implement more complex functionality computers are introduced as controllers As the computer is inherently discrete, the process can only be observed and controlled given instants in time This is not the case with analogue electronics or mechanical controllers RPM Engine HMI Computer Current Control theory Since the process is only observed at instants in time, the timing of the observations is important! Varying sampling intervals reduce control performance

3 Control theory The computer also introduces delays in the control signals! Feedback delays reduce control performance and may lead to instability Sample Actuate Engine RPM Delay Computer HMI Control theory To achieve constant sampling intervals and feed back delays in a computer control system is non trivial Since many activities share the same CPU, timing variations are inevitably introduced Hence, sampling periods will vary Feed back delay will be non zero and varying It is thus important to be able to control and analyse the timing of computer activities Which lead us to real-time systems and real-time systems theory Anti-slip (Traction Control) Detect derivates of wheel speed that do not adhere to car acceleration but rather to loss of grip. Reaction to loss of grip - reduce wheel speed Brake system traction control use brakes to reduce wheel speed Observe: wheel speed Actuate: brake force Every t seconds: readsensor(new_speed); diff(old_speed,new_speed); acc = diff/ t if(acc > limit) apply_breakes(); Wheel speed Computer Apply brake force 3

4 Time-triggered execution model 1. How fast do we have to process input and respond to the controlled process (delay)? 2. How often do we have to sample the environment to get a sufficiently good view of it. sampling actuation sampling actuation sampling actuation delay time Period time Period time Airbag Detect an impact Reaction inflate airbag Event: sensor impact detection Actuate: inflate airbag Impact Earliest time Latest time Event-triggered execution model Event occur CPU executes another task Event actuation delay time Response time: How long does it take for a task to respond (actuate) upon an event.

5 Why real-time systems theory? To fulfil timing requirements Predict timing behavoiur For inclusion of new functionality Achiving a robust temporal behaviour Trade-offs Resource usage Cost Functional requirements Other quality attributes maintainability, portability,... A C B D Computer Activities Dependencies Interactions 1. Specification Constraints 2. Translation A B D C What s a real-time system? A real-time systems is a system that reacts upon outside events and performs a function based on these and and gives a response within a certain time. Correctness of the function does not only depend correctness of the result, but also the timeliness of it. Real-time computer system The controlled process dictates the time scale (some processes have demand on response at second-level, others at milli- or even microsecond level). Sensors and actuators Sensors Observes application Not engineering rather e.g., voltage level A/D conversion Actuators Affects application D/A conversion Smart sensors and actuators Transformation to engineering Pre-processing of Lightweight networks Engine RPM Real-time computer system HMI Computer Current

6 Computer Nodes - Hardware CPU Memory Volatile RAM Persistent ROM Flash Digital in/out A/D and D/A converters communication ASICS Communication CPU Memory Real-time computer system Computer Nodes - Software Run-time system The execution engine, e.g., a RTOS Supporting software Networking Application software Real-time computer system Application Supporting software Run-time system Hardware Conceptual computer system architecture Sensors Actuators Computer Nodes Network Additional networks Gateways

7 An embedded system architecture XC90 ~40 nodes UEM PDM SUM SUB SCM MMS MP1 SHM AUD DEM ATM TCM RSM PHM SRM MP2 MMM CCM ICM GSM ECM ICM SRS BCM CPM DIM SWS SAS SWM LSM CEM SHM PSM DDM BSC PAS AEM REM ISM Heterogenous systems The complex embedded system is a hetrogenous system Hence, the computer system must serve many different needs Safety critical functionality Time-triggerd periodic activities feedback control (anti-slip) Event-triggerd sporadic activities discrete control systems (airbag) Non-critical functionality Time-triggerd periodic activities feedback control (Automatic volume adjustment) Event-triggerd sporadic activities discrete control systems (lights, window control) Several functions in same system Foreseeing the timing behavior and availability of resources for a given function is non-trivial It is also non-trivial to foresee the resource need for an entire system We need models and analysis to predict timing behaviour and resource usage sampling actuation event actuation time Period time delay Response time

8 Modelling embedded real-time systems The purpose of the models is to provide An abstraction of the functionality Hides unnecessary details about software and hardware A base for resoning and analysis Timing and resource usage Design Mapping from requirements (functionality) to the model Time-triggered vs. Event-triggered Mapping from model to HW architecture Real-time Operating system (OS) Run-time scheduling Scheduling analysis methods Proof by construction or response time analysis Why is it so difficult to construct real-time systems? Problem areas How to define the priority for an activity? Overload (more activity than the system can handle) Deadlock Inconsistent Full queues Starvation How do we specify the constraints? How do we verify correctness? Communication Distributed real-time systems..