Temperature Control Unit—Modeling and Implementation of a Particle Filter on a Microcontroller

Round 1

Reviewer 1 Report

This paper presents scope of particle filter estimation algorithm in embedded system STM32F407VGT6. The paper is interesting. However, few comments and suggestions are given below.

1. For three states: keeping is used as a name for one state. Wouldn't it be better to replace it with "normal" instead of "keeping"?

2. It would better to explain D (O) and D (S) platforms before presenting their results.

3. In Figures (4) to (6), The value of u2 is not taken zero when u1 is taken as 90%. Why these values have been considered.? What will happen if u2 is taken as 0%.

4. Please replace section title "Summary" with "Conclusion" and present your findings in this section.

5. In line 2 of abstract, should not it be "was built" instead of "was build".

6. In section 4 title, "A" cannot be used with plural nouns like "research objects"

7. Please proofread read the whole paper for English editing.

Author Response

"This paper presents scope of particle filter estimation algorithm in embedded system STM32F407VGT6. The paper is interesting. However, few comments and suggestions are given below."

We thank the reviewer for the review, which hopefully helped us improve the paper content and presentation.

"1. For three states: keeping is used as a name for one state. Wouldn't it be better to replace it with "normal" instead of "keeping"?"

We thank the reviewer for the comment. We looked at all the state names again and changed ``keeping'' to ``normal'' at each point.

"2. It would better to explain D (O) and D (S) platforms before presenting their results."

We thank you for the hint. We moved the list of implementation details from the end of the paragraph to the part related to platforms description.

"3. In Figures (4) to (6), The value of u2 is not taken zero when u1 is taken as 90%. Why these values have been considered.? What will happen if u2 is taken as 0%."

Thank you for the question. The u2 value is never set to zero as it is signal related to the fan. The minimal PWM duty cycle is 30% due to below the value the fan is not working correctly. What is much more important, we observed the fan is very useful to mixed up the air in the chamber. Using values below 30% for u2 would have a negative impact on the quality of the experiments. We've added an detailed explanation of this issue in the article.

"4. Please replace section title "Summary" with "Conclusion" and present your findings in this section."

Thank you for this comment. We changed the chapter name to ''Conclusion'' as suggested. The chapter contains the most important conclusions.

"5. In line 2 of abstract, should not it be "was built" instead of "was build"."

Thank you for your hint. Indeed, we overlooked this grammatical error. Thank you for noticing and improving.

"6. In section 4 title, "A" cannot be used with plural nouns like "research objects""

Thank you very much for pointing out another language error. We have corrected the title of the indicated chapter.

"7. Please proofread read the whole paper for English editing."

Thank you for this comment. We thoroughly analyzed the linguistic and grammatical correctness of the text, introducing the necessary corrections. We hope that the language level of the work will now be noticeably higher.

Reviewer 2 Report

The article clearly expressed the usage of temperature control unit for microcontroller by the implementation of the particle filter estimation algorithm. The deferent operating conditions are verified and results were presented in well manner. The following observations will be considered, before the next process of publishing the work.  

 1.   Identified gaps in the existing temperature control units and its challenges how to improve by your Bootstrap Particle Filter design and implementation? Add in text form in the introduction section.

 2.   Platform D standard and optimal version simulation and controlling parameters could be add in  a table form.  

 3.   Adding the flow chart for the step by step implementation of your proposed algorithm, will be give good understanding to the readers.  

Author Response

"The article clearly expressed the usage of temperature control unit for microcontroller by the implementation of the particle filter estimation algorithm. The deferent operating conditions are verified and results were presented in well manner. The following observations will be considered, before the next process of publishing the work."

We thank the reviewer for analysis of our paper, which hopefully helped us improve the paper content and presentation.

"1.   Identified gaps in the existing temperature control units and its challenges how to improve by your Bootstrap Particle Filter design and implementation? Add in text form in the introduction section."

We thank the reviewer for the comment. We analyzed the Introduction section and the literature again. We expanded the description in the Introduction section as suggested. We also cited a few more papers on temperature control.

" 2.   Platform D standard and optimal version simulation and controlling parameters could be add in  a table form. "

Thank you for the hints. We have added an additional table as suggested. The table contains simulation parameters (such as number of iterations and number of particles) and corresponding to them execution times in milliseconds, respectively for standard and optimized version.

"3.   Adding the flow chart for the step by step implementation of your proposed algorithm, will be give good understanding to the readers.  "

Thank you for this comment. In the ``Particle filter'' chapter, we have added a flow chart that shows exactly how the algorithm works. The flow chart describes in detail the operating steps of the systematic resampling. We hope it will help readers to understand our implementation.

Reviewer 3 Report

Dear Authors,

Your paper presents a reiteration of the work of N.J. Gordon, Salmond, and Smith (1993) with the addition of the control algorithm ADRC instead of PID of Wu et al. (2021) and also your work in On transitioning from PID to ADRC in thermal power plants(2021). Also, the experimental work added to the paper has recentered the purpose on applications of particle filtering/estimation towards low energy consumption using low-cost boards. 

The paper is written in a direct manner. It brings a sufficient mathematical description. The experimental setup and work are well presented.

However, there are some aspects that can be considered for the final manuscript. I present them in the following enumeration:

- Which is more appropriate or correct? (Table 1) Bootstrap Particle Filter or Bayesian Bootstrap Particle (see 1993 Gordon et al work)? 

- in table 2 FPGA term was introduced but not used in the text. If FPGA term is relevant to the manuscript please add a paragraph for readability.

- the same issue as before in table 2 the term GPU. Please provide a paragraph for readability and relevance.

- a discussion regarding the selection of the number of particles and the average execution time for an application. Your experiment was of a small scale and the experimental environment was controlled. Can the experiment be conducted in direct interaction with the environment? How it should behave? I recommend that you add limits to the design of your experiment, especially when you plan to improve quality of the measurements performance by UAV on-board sensors in the future.

- Figures 4,5 6 can be much more detailed. Add captioning for the colors used in the legend.

- A combined figure that includes figure 4, 5 and 6 could elevate the results presented in 258-269 lines.

- also,  a paragraph for reproducibility can be added to Chapter 3. to describe the box enclosure, dimensions.  Figure 2 can be remade with text and arrows to describe presented components in the picture

- Also the experiment might be sensible to air pressure. Simultaneous temperature measurements for both sensors should be included in 4, 5, 6.. This is due to external ambiental temperature and the fact that the box exchanges heat with the ambient. Please add a paragraph that includes your considerations in the experiments  including physical ones and if possible include the reference temperature for each experiment regarding number o particles considered.

- a typo error is at line 186

I look forward to see the final paper.

Kind regards

Comments for author File: Comments.pdf

Author Response

"Dear Authors,
Your paper presents a reiteration of the work of N.J. Gordon, Salmond, and Smith (1993) with the addition of the control algorithm ADRC instead of PID of Wu et al. (2021) and also your work in On transitioning from PID to ADRC in thermal power plants(2021). Also, the experimental work added to the paper has recentered the purpose on applications of particle filtering/estimation towards low energy consumption using low-cost boards.
The paper is written in a direct manner. It brings a sufficient description. The experimental setup and work are well presented.
However, there are some aspects that can be considered for the final manuscript. I present them in the following enumeration:"

Thank you for the review. The comments allowed us to improve the article content and presentation.

"- Which is more appropriate or correct? (Table 1) Bootstrap Particle Filter or Bayesian Bootstrap Particle (see 1993 Gordon et al work)? "

The algorithm proposed by Gordon et al. was Bayesian Bootstrap Filter or shorter - Bootstrap Filter (in abstract, as well as in the text). In the mentioned article authors do not use phrase ,,particle filter'', because they probably did not know that thanks to their work this kind of filters would be created. Moreover - they do not use even ,,particles'' word, and use ,,samples'' instead, because proposed Bootstrap Filter was a modification of the SIS (Sequential Importance Sampling) algorithm, in which ,,samples'' are used, not ,,particles'' (but actually they are the same). Later, it was assumed that particle filters (or SIR - Sequential Importance Resampling algorithms) differs from the SIS algorithms only in that they have the Resampling step. Hence, the Bootstrap Filter was the first particle filter, because Gordon, Salmond and Smith proposed usage of the resampling.

All the next particle filters have ,,particle filter'' in their name, e.g. Auxiliary PF (Pitt et al. 1999), Rao-Blackwelised PF (Doucet et al. 2000), Extended Kalman PF (Merwe et al. 2000), Unscented PF (Merwe et al. 2000), Lattice PF, Regularized PF, Likelihood PF, Gaussian PF, Self Adaptive PF, Twisted PF, and many other. The only exceptions are SMC (Sequential Monte Carlo) algorithms, where ,,Monte Carlo'' phrase occurs in their name, e.g. Marcov Chain MC (Freitas et al. 2001) or Population MC (and these algorithms are still particle filters).

Therefore, if in every particle filter algorithm there is ,,Particle Filter'' (or ,,Monte Carlo'') phrase in its name, we extend ,,Bootstrap Filter'' too ,,Bootstrap Particle Filter''. In the current article it gives no difference, but in articles where different types of algorithms are compared, it is easier for readers to associate BPF with PFs algorithms than BF algorithm. And we try to unificate the designations and descriptions in our articles, regardless of the current content.

"- in table 2 FPGA term was introduced but not used in the text. If FPGA term is relevant to the manuscript please add a paragraph for readability."

Thank you for the comment. Several abbreviations were not related to the article, which we missed. We rechecked the document again and removed unused phrases.

"- the same issue as before in table 2 the term GPU. Please provide a paragraph for readability and relevance."

Thank you for the comment. The answer for the previous suggestion seems to be appropriate also here.

"- a discussion regarding the selection of the number of particles and the average execution time for an application. Your experiment was of a small scale and the experimental environment was controlled. Can the experiment be conducted in direct interaction with the environment? How it should behave? I recommend that you add limits to the design of your experiment, especially when you plan to improve quality of the measurements performance by UAV on-board sensors in the future."

We thank you for this comment. In the future, it is planned to carry out tests under various environmental conditions, for example at very low or very high outside temperatures. We have described this in the Conclusion section as further work development prospects (e.g. by placing the box in an oven or fridge).

"- Figures 4,5 6 can be much more detailed. Add captioning for the colors used in the legend."

Thank you for this comment. We have updated the captions to the indicated charts. We have marked how the individual graphs are presented to facilitate their interpretation.

"- A combined figure that includes figure 4, 5 and 6 could elevate the results presented in 258-269 lines."

Thank you for this comment. We have added a new figure that shows the results of the above-mentioned combined together. We presented there graphs of estimation errors for the three shown tests.

"- also,  a paragraph for reproducibility can be added to Chapter 3. to describe the box enclosure, dimensions.  Figure 2 can be remade with text and arrows to describe presented components in the picture"

Thank you for the suggestion. We have expanded the section by giving additional figure and modifying existing one. We also highlight the outer dimensions in the text.

"- Also the experiment might be sensible to air pressure. Simultaneous temperature measurements for both sensors should be included in 4, 5, 6.. This is due to external ambiental temperature and the fact that the box exchanges heat with the ambient. Please add a paragraph that includes your considerations in the experiments  including physical ones and if possible include the reference temperature for each experiment regarding number o particles considered."

Thank you for this comment. We have also included the external temperature waveforms in all the graphs. We also added conclusions regarding the influence of external temperature on the operation of the system in the text.

"- a typo error is at line 186"

Thank you for your comment and for noticing the typo. We corrected the ``state-space'' expression.

Round 2

Reviewer 1 Report

The quality of manuscript is considerably improved. It can be accepted for publication.