You are on page 1of 8

ACS11001

Feedback Coursework: General feedback was provided to class before the written exam. In
addition, students can view comments annotated within their reports on turnitin.

Exam: Q1 - most students did well on this question. Just minor mistakes e.g.
missing minus sign for a negative number.
Q2 - this was the most challenging question as it was on the full-adder circuit
with a twist on the inputs which were given as functions instead of the typical
input variables that students had seen in class/tutorial. But quite a number of
students still scored good marks of above 6/10. It needed working out the values
of the given functions (can be easily done using a truth table) which then
becomes the inputs into the full-adder. The rest follows the same approach for
implementing multiplexers as seen in lectures.
Q3 - should have been a straightforward question as no calculations were
needed, but just definitions from textbooks. Perhaps some students ran out of
time? or they missed out studying the topic as it was the very last topic covered
in lectures?

ACS219

Feedback In general, the students did well in quizzes and lab. The final assignment was a
group activity with peer assessment which differentiate between marks. I would
expect more tutorial attendance from this cohort which could help them doing a
better exam.

ACS317

Feedback Coursework: Feedback already provided to students.

Exam: I must admit to being very disappointed in the performance of the


students on this exam. Reading the answers I fear students focussed on
memorisation of techniques rather than understanding so any questions probing
their understanding, even at a very basic level, were largely unanswered,
irrespective of how easy the required answer was intended to be.

All exams will have some routine questions which are straightforward
reproduction of standard methods and this exam was no different. Many of the
questions were intended to be 'easy' marks so students demonstrated basic skills
and understanding. However it seems that the simple rewording using words like
'analyse' so students had demonstrate comprehension and understanding to
choose what method or analysis tool to use rather than leading by the nose, e.g.
'by finding the eigenvalues/controllability matrix/etc.', many students were
unable to progress.

Detailed analysis of each question

Q1a: I had expected that the deriving the model was easy marks, and then
finding the state model was a soft introduction to the exam. I assumed most
would get full marks here. This type of question appears every year so nothing
novel and indeed the modelling equations here are much easier than many from
past exams. The main difference is that previous exams have often used
elaborate scenarios and given the differential equations. Here I used a simple
scenario so that students could demonstrate the full process, modelling through
to the SS model.

I was flabbergasted that only a small minority were able to write down the
equation "A(dh/dt) = fin - fout" for a tank level system and thus failed question 1
entirely. This is year 1, semester 1 material and also covered in lecture 1 of
ACS317 and surely any graduating student taking ACS317 should be expected to
be able to do this level of modelling, especially in a module whose learning
outcomes include derivation of state space models.

Q1b: While the numerical part used the result from part Q1a, students could still
have got more than half the marks available by showing they knew how to form a
discrete model from a continuous one (a common past exam question) and yet
most chose not to do this. There is nothing novel, difficult or different about this
question compared to previous exams.

Q2: Most students could write the canonical form from memory which gained a
few marks, but pretty much no-one showed the requested detailed explanation
of why this works. We cannot give high marks just for memorisation.

Q3: While most students did this question fine and got near to full marks, a few
seemed totally baffled and yet this is a standard question from most past exams
and we did numerous times in lectures, including the revision lectures based on
past exams of week 11. The only difference in question presentation is that
students would not told to use the transformation z = Tx but rather required to
propose this themselves (e.g. not led by the nose).

Q4: This question covers the fundamentals of SS behaviours and was open-ended
(essay type) so that students could demonstrate what they knew and thus gain
marks. Again, it seems many students were not comfortable with this and largely
students did very little here beyond computing the eigenvalues. Astonishingly
several did not notice the model given was discrete and defined the poles as
unstable when they were inside the unit circle. Several students were computing
eigenvalues and vectors by hand despite having MATLAB!

Q5: This was routine bookwork (essay type) on feedback, pole placement, LQR
and observers and this type of question appears in every exam, albeit with
different foci each year. Most students gained reasonable marks on the LQR and
integral control discussion but very few made the link between full state
feedback and the need for an observer and thus did not answer part (c). Again an
example where, because the question did not 'lead students by the nose' and tell
them explicitly to discuss observers, many totally failed.

Q6: This question includes parts which were designed to be challenging so that
the very best students could be differentiated from the others. Having said that,
more than half the marks were still straightforward.

a) This appears in every past exam. Students simply needed to explain the
process of linearisation (in abstract). Despite this being a common exam
question, very few answered.

b) This was the harder bit, apply the process from (a) to the given equations.
Students could still have gained good marks by showing intent and organisation,
but many left this blank.

c) Meant to be easy marks requiring standard analysis of the model provided.


Well done in the main, but many failed (did nothing) because they did not link
the word analyse to doing things such as finding eigenvalues and assessing
controllability and observability.

d) Students asked to incorporate integral action and design an LQR control (can
use lqr.m for number crunching). This is a standard question from past exams,
and covered extensively in lectures at the very end of the course, but most did
nothing here. The simulink part carried just a few marks and students had been
advised throughout term of exactly what simulations would be expected and
advised to prepare a suitable file (templates and tutorial support provided). Very
few seem to have done this.

e) Students asked to develop an observer using pole placement. Again the


conceptual side is a standard examination question in past papers and they could
use MATLAB for number crunching, but few wrote anything at all.

The simulink part, as above, carried a few additional marks - almost no-one did
this.
ACS318

Feedback Exam: Candidates had a distribution of marks with four reaching marks of 90 or
greater and two fails. 45% achieved a mark of 70 or above. Amongst the three
questions that had to be answered, students performed the best in Q2 answering
part a) correctly and showing evidence of understanding in part c). Where
students didn’t appear to recall the method. Next popular was Q1 where most
answered parts a) and e) but with some mixed performance in the other parts,
often leaving the answers with just expressions without explanations. Part e) also
lacked brief discussion on design parameters that appear to have been
overlooked. Students appear to have struggled with Q3, most notably in part b).
Part a) answers were sometimes incomplete and some lost marks in part c) for
not recognising that this is an output only model. Overall, the performance was
strong with a mean mark of 65.

ACS323

Feedback Coursework: Students did generally well in the coursework.

Exam: Students performed rather better in the exam component as compared to


the coursework, perhaps due to the fact that the exam took place later in the
period (01/02/2023), which meant they had time to prepare well- Most students
attempted both questions enthusiastically and they showed that they
understood many components of the module.

ACS327

Feedback Exam: In general, the performance for the ACS327 exam was very good. I would
like to mention a few errors that appear in works more often than I expected.

1. In Question 1b i) students were asked to derive an equation for the effect of


atmospheric drag on the rate of change of orbital radius. In some scripts this
relation was provided without derivation and in some other scripts the relation
for the on the rate of change of orbital period was provided instead.

2. in question 2d the change of orbital plane manoeuvre was required to transfer


the satellite to GEO. Some scripts omitted this manoeuvre. In other scripts
instead of the optimal combined manoeuvre of plane change and circularisation
two separate manoeuvres were presented.

3. Methods and/or miscalculations in applications of these methods to obtain


required answers were frequent problems with question 3a.
ACS329

Feedback Coursework: Students put a lot of effort into completing the assignment for this
module's practical part which was reflected in their marks. Most submitted
assignments were of the high quality.

Exam: Exam marks were lower than expected with a very wide range of marks
between the students. Most students did well on the first question, but the
second and third questions created a large division between the marks. It was
also clear from the more narrative questions that some students did very little
revision of the lecture material. Some answers were, on the other hand, of
exceptional quality.

ACS6126

Feedback Exam: In general, students performed very well during this exam. One of the
errors I would like to mention is that instead of a charged particle trajectory in
the magnetic field (q1), scripts present derivations of a drift motion in drift
approximation.

ACS6128

Feedback Coursework: Coursework comprises two assignments, accounting for 20% of the
final module grade. Students showed very good engagements in coursework. The
overall performance on coursework is excellent.

Exam: The final exam comprises four questions, each is marked out of 20 marks.
The exam accounts for 80% of the final module marks.

The majority of the students showed very good performance, leading to a good
average. A few students, who had a lack of engagement in coursework and who
had not performing well in the exam, did not show a satisfactory overall
performance.
ACS6129

Feedback
Coursework: Coursework comprises two assignments, accounting for 20% of the
final module grade. Students showed very good engagements in coursework. The
overall performance on coursework is excellent. Well done.

Exam: The majority of the students showed very good performance, leading to a
good average. There are some excellent students (over 90) and some struggling
students (under 40). Most students performed well on Q1b, Q2b. Q2c, Q3. Q1d
and Q2a are a bit challenge to some of the students.

ACS6408

Feedback Coursework: Coursework comprises three assignments, each accounts for 10% of
the final module marks. The majority of students have been highly engaged in
coursework, but some lacked effort and attitude. Overall, most students
performed very well on coursework, which was evidenced by the relatively high
average.

Exam: The final exam comprises three questions, and accounts for 70% of the
final module marks. The majority of the students showed very strong
performance in the exam, leading to a very good average.

ACS6427

Feedback
Coursework: I was pleased to see excellent engagement throughout the semester
with the Blackboard quizzes. I was particularly impressed by the diversity of
methods that students attempted and validated in the MATLAB coursework.

Exam: Great job to the students who attempted Question 1 on linear regression
and Question 4 on PCA as they mostly achieved good marks! Some students
experienced difficulty recalling how to derive the OLS regularization analytical
solution required in Question 2. Although Question 3 had the lowest overall
scores, this presents a clear area of improvement for students to focus on in their
future studies. It highlights that many students should further develop their
calculus skills especially when it comes to calculating gradients, required in this
question to iterate the gradient descent algorithm.
ACS6501

Feedback Coursework: Feedback already provided to students.

Exam: Marks range from 21 to 91. Almost all of you did exceptionally well on a
number of core topics including complex numbers, Jacobian matrices, and the
row reduction method. Some of you had difficulties dealing with mathematical
notation, for example, where used to define sets: I suggest reading the two pages
on Set Theory (Section 0.3) of our core textbook (Hubbard & Hubbard). Being
able to confidently handle mathematical expressions should prove useful in your
dissertation project. It is often good practice to formally define the problem you
are working on (e.g. in a dedicated "Problem Formulation" section). Some of you
dealt comfortably with the questions regarding eigenvectors/eigenvalues,
whereas others had difficulties deriving the requested relations. Think about the
approach of finding solutions that we practiced in class. It often requires you to
recall some core equation (in this instance A * x = lambda * x) and then you need
to think what the equation should look like once suitable modifications have
been made (e.g. A * A * x = lambda * lambda * x). Next, you can think of what
modifications will bring you there. Overall, I am proud of the progress you have
made in Semester 1. If you encounter some maths obstacles in Semester 2 or
your future job - and I hope you will - I recommend consulting again the textbook
we used (Hubbard & Hubbard), as you should now have the skills to do so more
independently :-)

ACS6503

Feedback Exam:

Looking at Q1, 40 students (69%) did not attempt at least one part of the
question. The answers to this question were mainly found in the safety section of
the lectures. In these risk and hazard were defined as well as the standard
process for risk reduction for any robotic system.

Q2 was generally well answered, although care needed to be taken over getting
the frame rotations correct.

In Q3 there were many incorrect versions of the DH used to find the transform
for the robot. In this case there were 4 rows each starting at a joint and
consisting of a joint and link, l1, j1 & l2, j2 & l3 and j3 to give four rows. Often the
fourth row was added to the third row, this moves j3 behind l3 in the structure
and is incorrect. The frame rotation had to be chosen to select an orientation
that meant to translation or rotation in y-axis.

ACS61013

Feedback Feedback already provided to students.


ACS61015

Feedback Feedback already provided to students.

You might also like