2IN27 : Quantitative Evaluation of Embedded Systems 2013/2014


This is the primitive website that was used for the QEES course in 2013/2014. The new edition of QEES 2014/2015 will be given using a SAKAI site including forum, news o and hand-in possibilities for assignments. If you would like to follow this edition of the course and have not been registered automatically yet, please email me..
This webpage is no longer maintained...

Announcements

PRACTICE EXAM NOW ONLINE!

Don't forget to take a look at the exercises already.

14-nov-2013 : Thanks to Robin Wolffensperger and Rubin Lubben for pointing out an error in Weblecture 1.
The simulations I do in fast-forward are not entirely correct.

15-nov-2013 : Today it turned out that making and discussing exercises takes more time than I expected.
I did not perceive this as a bad thing, but I am very curious to know how you experienced todays session.
Please drop me an email if you have anything to comment!

Next Monday we will start with the material I actually planned for today.
After that, I plan to discuss graph 7 from the exercises with you in detail.
Especially determining the optimized periodic schedule as a function of the period is tricky for this one,
so you might want to try it already at home!

17-nov-2013 : Thanks to Dimitrios Chronopoulos for pointing out an error in Weblecture 4.
At the end, when I optimized the latency, I forgot to take the newly introduced arrow into account.

General thanks to Tim Kers, Femke Jansen and Marijn Blom, en Janina Torbecke for pointing out some smaller errors and for several remarks on how to make this course better.

18-nov-2013 : For the recordings of last-years lectures on Dataflow, see the TU/e videolectures site.
You will find the lectures under the computer science tab at 2IN27.
This years lectures (the Dataflow part) should be put there as well, but the processing seems to lag behind.
I will try to obtain those videos and put them on this site as well, but this process is largely out of my control.

20-nov-2013 : This Friday I'll discuss the assignment and run a tutorial on the SDF3 toolset. Please bring your laptops! or make sure your neighbour does ;-) If you want to install the tutorial ahead of time, click this link.

23-nov-2013 : Thanks to Marijn Blom and Ivar in 't Veen for pointing out that I accidentally put the wrong version of "example 3" in the tutorial. The zip file below has been updated, and here is the example in its original form.

25-nov-2013 : I have received a number of requests for 'a book on dataflow to read'. Unfortunately, there are not that many books available. However, today I have added a link to the PhD-thesis of Orlando Moreira. Chapter 3 and 4 of this thesis are about dataflow and periodic scheduling and should be informative. If you decide to study those, please tell me if you think it is suitable course material.

28-nov-2013 : Unfortunately Arjan Mooij is ill, so the guest-lecture tomorrow has to be postponed. Instead Marielle Stoelinga will just continue her series about timed automata.

30-nov-2013 : Today I updated the dataflow assignment a little bit (removing some subquestions) because I thought is was too much work. Furthermore, I put a template online of what I expect you to hand in.

2-dec-2013 : Regarding the SDF3 tutorial. Exercise 4 may give an error message that you can ignore, and the output will be a one-line xml file which should open fine in any browser, but will be difficult to read in, for example notepad.

3-dec-2013 : Today I managed to include the recordings of the class lectures as well. I will add the most recent lectures as soon as I get hold of the proper link to embed. Building a webpage with embedded videos is new to me, and while youtube was making it easy for a novice html programmer, using the TU/e streaming server is a bit more complex. So if you have any trouble using your particular browser, let me know. I cannot promise to fix it immediately, but I will do my best to learn how to fix it... (I already noticed I cannot stream to my Android tablet. I'm affraid this is something I can only fix in next years edition of this course.)

4-dec-2013 : TIP regarding the dataflow assignment. Some students get the error message "Failed to load SADF graph". In that case, check whether you accidentally corrupted the xml format by opening your xml file in, for example, a browser or editor that parses xml. Also, some students get the error message "Graph deadlocked". In this case there may be something wrong with the rates you have chosen, or with the number of initial tokens. This can also be caused by mundane typo's, because misspelled commands are simply ignored by the tool (typing initialToken = '5' will simply leave the initial tokens at 0, because the command is initialTokens).

5-dec-2013 : Thanks to Oscar Starink and Dimitrios Chronopoulos for jointly making me realize a typo in the TDMA model. The latency is P-S, the execution of a q-th piece of slice takes S/q, and consequently, that piece is delayed by P-(S/q). The slides have been updated accordingly, the CPS model is only affected in the computations (not updated), because in the case of hops we have q=1 anyway.

5-dec-2013 : There is a difference between the rates for the computations suggested by the picture of the CPS-model dataflow graph (1,2 and 2, respectively) and the actual xml graph (1,3 and 7). Use the ones in the xml graph as a starting point for your case study. The ones in the picture were typo's.

5-dec-2013 : To improve the page load-time, I've replaced most of the embedded movies by links.

16-dec-2013 : The template for the second assignment has been posted. The deadline for this has been postponed until 8th of January, but be aware that Marielle Stoelinga will not be available for questions during the Christmas break. Note that question 4c has been removed.

16-dec-2013 : Note that the paper about datapath modeling in OCe is considered exam-material for this course.

20-dec-2013 : Results for the first assignment will become available after the holiday (i.e. on the 6th). Just to be clear for those who fear a negative judgement: after receiving feedback you will be expected to hand in proper results for the same assignment afterwards. I am not going to extend the exercise or anything like that.

20-dec-2013 : As it turns out, there are a number of unclarities in the second assignment on timed automata. Since Marielle Stoelinga is currently unable to resolve this, and unavailable for questions during the holidays, we have decided to make this assignment optional rather than mandatory. If you hand it in and it is graded as a pass you will receive a 0.5 point bonus (out of 10) on the final exam.

6-jan-2014 : To all Eindhoven students: note that the lecture of this Friday will take place at the usual place and time, contrary to what you might have found in the schedule on OASE.

6-jan-2014 : Everyone should have received an email with the result of assignment 1 now. If you did not, contact me. Perhaps I used a different email adress. (Some Delft students turn out to have a TU/e email account as well, so check that too...)

9-jan-2014 : The set of solutions of the pCTL exercises has been updated.

15-jan-2014 : Some of you may have noticed that the last link of the video-lectures is not working yet. The link is not broken. The video has simply not yet been uploaded by the people responsible for the recording. Sadly, this is out of my control, but I apologize for the inconvenience.

15-jan-2014 : As was announced during the question hour and the previous lecture by Anne Remke, it is allowed to bring one hand-written A4 double-sided with you on the exam.

17-feb-2014 : Dear Students,

By now, you should have received your final score for the QEES exam. In this email, we will briefly explain how this score has been calculated, because we received some concerns of students with regards to two matters: (i) the length of the exam (ii) the availability of previous exams.

(i) Length of exam. The exam results clearly showed a drop in the expected scores towards the end of the exam, which leads us to conclude that the complaints about the length of the exam may be justified. Therefore, to compensate for the length of the exam, ALL grades were increased by 6%. With this modification, the passing rate of the exam was 69.5%
(ii) Availability of practice exams. Without our knowledge, practice exams of previous years turned out to be available in Eindhoven. This might have given the Eindhoven students an advantage. However, even for question 3 (which was quite similar to the question of a previous year), the maximum difference among the average grade of the three institutions was 1.5 out of the 10 points assigned to Q3. Regarding the final grade, the maximum average difference was 0.38 out of of 10. These are not meaningful statistical differences (< 4%), hence no remedial actions were taken here.

If you have any questions regarding the exam or other aspects of QEES, please let us know.

Best Regards,

Anne Remke, Pieter Cuijpers, Marco Zuniga and Marielle Stoelinga

28-mar-2014: We have now published the previous exam with a set of solutions, so you can practice for the upcoming resit. You can find it here. Good luck.

Before we start...

Before we start with the real course... Anne Remke interviewed three prominent researchers about QEES: Kim Larssen (the founder of the tool UPPAAL), Edward Lee (one of the inventors of synchronous dataflow), and Holger Hermanns (a very well-known specialist in probabilistic and hybrid systems). Watch the interviews here!

Dataflow (info: Pieter Cuijpers)

Monday 11-11-'13

Recording of the first class lecture
Slides of the introduction to the course, including the course program
Slides on max-plus algebra

Weblecture 1 on Dataflow
Slides weblecture 1

Friday 15-11-'13

Weblecture 2 on Dataflow
Slides weblecture 2

Weblecture 3 on Dataflow
Slides weblecture 3

Recording of the second class lecture
Slides with scribbles on the exercises

Monday 18-11-'13

Weblecture 4 on Dataflow
Slides weblecture 4

Weblecture 5 on Dataflow
Slides weblecture 5

Recording of the third class lecture
Slides of the third class lecture

Friday 22-11-'13

Weblecture 6 on Dataflow
Slides weblecture 6

Weblecture 7 on Dataflow
Slides weblecture 7

Recording of the fourth class lecture
Slides of the fourth class lecture on TDMA and the CPS assignment

Case study : deadline 9-12-'13

This years assignment is a case study to the use of several quantitative evaluation methods for the analysis of a cyber-physical control system. The first part of this assignment consists of a tutorial regarding the SDF3 toolset, and the actual assignment.
You will find the tutorial here. Install it by unfolding it in a convenient place, opening the index.html using a web-browser, and following the instructions.
Furthermore, to make things easier for you, here you find a rich-text file containing a template of what I expect you to hand in for this assignment.

Just to be completely clear. You are allowed to make the assignment in pairs, but you should each send an email to me separately.

Literature and Additional Exercises

I have compiled a set of basic exercises using which you can prepare yourself for the exam. And, upon many requests, a set of solutions (just the final answers) for some of the graphs: Graph 1, Graph 3, Graph 4, Graph 5, Graph 6, Graph 7

For further reading on dataflow, I advice the following books and papers:
E. Lee and D. Messerschmitt. Synchronous data flow. Proceedings of the IEEE, 1987
S.Sriram and S. Bhattacharyya. Embedded Multiprocessors: Scheduling and Synchronization., Marcel Dekker Inc., 2000
F.Baccelli, G. Cohen, G. Olsder, and J.-P. Quadrat. Synchronization and Linearity. John Wildey & Sons, Inc., 1992
Free online version
O. Moreira Temporal Analysis and Scheduling of Hard Real-Time Radios on a Multi-processor PhD Thesis, TU/Eindhoven, 2012
TU/e archive
(Section 3 and 4)
A. Lele, O. Moreira, P.J.L. Cuijpers. Dataflow Based Temporal Analysis for TDM. The International Conference on Embedded Software (EMSOFT-2012)

Timed Automata (info: Marielle Stoelinga)

Monday 25-11-'13

Before the first class-room lecture on timed automata, you should study these three power-point presentations.
Each slide contains an audio-plugin that explains the slide.
The class on Monday will start with a discussion of the two exercises mentioned at the end of the lectures, which we expect you to prepare.
If you understood the lectures, this should be just 5-10 minutes work.
Screencast 1 on Timed Automata
Slides screencast 1

Screencast 2 on Timed Automata
Slides screencast 2

Screencast 3 on Timed Automata
Slides screencast 3

Recording of the first class lecture
Slides of the 1st class lecture.

Friday 29-11-'13

Screencast 4 on Timed Automata
Slides screencast 4
Recording of the second class lecture
Slides of the 2nd class lecture.

Monday 2-12-'13

Some exercises to prepare yourself for the class lecture and their solutions .

Recording of the third class lecture
Slides of the 3d class lecture.

Friday 6-12-'13

Short class lecture on regions
Screencast 5 on Timed Automata
Slides of screencast 5 5

Some exercises to prepare yourself for the exercise class.
The files Jobshop2.xml and Jobshop.q needed in the assignment.
And a final set of exercises. with solutions.

Monday 9-12-'13

GUEST LECTURE by Arjen Mooij on Model Based Design in Industrial Practice

Case study : deadline 8-1-'14 (this assignment is now optional, see announcement section)

The integral assignment about CPS and Timed Automata
A template for handing in the assignment
Anne Remke's paper about Wireless HART.

Literature and Additional Exercises

UPPAAL tutorial by Frits Vaandrager.
UPPAAL tutorial by Behrman, David and Larssen
A formal introduction to timed automata by Bengtsson and Yi

Furthermore, note that the paper about datapath modeling in OCe is considered exam-material for this course.

Markov Chains (info: Anne Remke)

Friday 13-12-'13

Class lecture on DTMC's
Slides on DTMC's

Monday 16-12-'13

Screencast 1 on CTL
Screencast 2 on CTL

Slides screencast 1
Slides screencast 2
Exercises on CTL and their solutions .
Recording of the exercise lecture.

Friday 20-12-'13

Screencast 3 on pCTL
Screencast 4 on pCTL

Slides screencast 3
Slides screencast 4
Exercises on pCTL and an updated version of their solutions .
Slides of the class lecture.
Recording of the exercise lecture.

Monday 6-1-'14

Screencast 5 on CSL
Screencast 6 on CSL

Slides of screencast 5
Slides of screencast 6

Slides of today's lecture
Slides on Wireless Hart
Class lecture on CTMC's.

Friday 10-1-'14

Exercises on CSL and their solutions .
Recording of the exercise lecture.

Case study : deadline 17-1-'14

The integral assignment about CPS and Markov Chains
Like with the previous assignments, it is allowed to hand in your results in pairs (at most tripples). One email per pair will do.

Literature and Additional Exercises

Paper on DTMC's by Boudewijn Haverkort
Tutorial on DTMC model checking by Christel Baier et all.

Question and answer hours

Monday 13-1-'14

Pieter Cuijpers and Stefano Schivo will be present to answer questions about dataflow and timed automata. For questions about markov chains you can email Anne Remke or our QEES3TU email adress.

Here you find the region automaton (solution to TA exercise 4 of assignment 3) that was requested during the question hour. The clock conditions between parentheses (e.g. x - y > 2) were added to show why location l0 cannot be reached from location l1.

Additional study material

The exam of january 2014 with solutions