26 May 2017
I’ve been gravitating more towards working on electronics projects. It’s a nice change from pecking the keyboard all day and it has plenty of interesting challenges of its own. The biggest difference I found is that you just can’t take your hardware working correctly for granted. Things break and catch fire all too easily if you’re not careful with what you’re doing.
My plan is to make a quadcopter from scratch (a rather loose term) and give it some ability to navigate unsupervised. This approach is (aptly?) called Simultaneous Localisation and Mapping, or SLAM. SLAM is by no means a cheap affair, be it in terms of hardware or software complexity. Using SLAM outdoors normally requires some kind of 3D localiser (such as a depth camera or LIDAR), GPS, and a fairly beefy computer to map the environment.
Of course, I’m setting my expectations low, and I will be very happy if I can get something out of this which is 10% as good as a human flying quadcopter by hand.
It all begins with the hardware and I’m going with this parts list:
- 1045 Propellers x4
- REES52 DJI F450 Frame
- REES52 CC3D F1 Flight Computer
- REES52 1000KV A2212 30A Brushless Motor x4
- REES52 SimonK 30A Electronic Speed Controllers
- SunRobotics 2200mAh 3S 35C LiPo Battery
- FlySky FSCT6B Computer Transmitter
The amount of embedded systems tech in these components is astounding. Each motor is driven by an ESC, which is essentially a full-fledged 8-bit SoC. The CC3D Flight Computer is an engineering marvel in it’s own right - it comes with a 32-bit/72MHz flashable microcontroller, an MPU6000 gyroscope/accelerometer for the IMU, and has expansion ports for GPS and Telemetry data. Even the battery is remarkable - it’s capable of discharging 850W at peak performance.
What I’m interested in, however, is controlling the CC3D in real-time using an Arduino. Typically, the CC3D is connected to a radio receiver which is wirelessly linked to a transmitter remote which is controlled by hand. The reciever is interfaced to the CC3D by outputting a PWM to each of the 6 channels when the transmitter is operated. Each channel is usually dedicated to only one axis of movement. For example, there are four dedicated channels for throttle control, yaw, roll, and pitch.
The idea behind this is that one day, I will have a lightweight computer (anyone willing to donate an NUC?) on the quadcopter doing SLAM in real-time. The computer will send throttle, pitch, yaw, and roll commands to the Arduino. The Arduino in turn will feed the CC3D with control signals to execute the maneouver plan. It all may seem like a Rube Goldberg machine, but it’s a lot easier than building my own flight computer to do the same.
To kick things off, I had to find the PWM frequency the receiver uses for controlling the ESC. To do this, I connected a pin from the output channel of the receiver to an ad-hoc Arduino Leonardo oscilloscope. Using the Processing library, we can get a 0-5V reading of the value on the A0 analog input of the Arduino.
Not a bad start! The peaks are a pretty clear sign that it’s a PWM with a low duty cycle. Using a ruler and some more code, I found that the peak to peak time is about 20ms, which works out to be a PWM frequency of 50Hz. This is in line with some websites speculating it works similarly to a servo control.
The length of time at the peak gives us the signal we need to send using the Arduino. Using some more analysis of the waveform, it looked like the signal were about 1-2ms long. A bit more research yielded that this is very similar to the waveform used for controlling servos. Fortunately, the Arduino has a Servo library meant just for this purpose. Using the writeMicroseconds() function, I got a pretty similar waveform:
I would go as far to say that this is an even cleaner signal than what the reciever outputs. I needed a 5V->3.3V voltage divider for the output, so I used 3 identical resistors in series to do the job.
Next thing was to get the CC3D to accept the input. For this it requires manual calibration of the transmitter. In Librepilot, the procedure is quite straightforward - you need to shift the sticks through their minimum, maximum, and neutral outputs to calibrate the CC3D for your transmitter. To do this, I mounted a push button on the Arduino which would cycle through 3 states each time it is pressed - 1100uS, 1500uS, and 1900uS. Not very elegant. Then again, if it’s stupid but works, it ain’t stupid.
I will be building the rest of the quadcopter once the parts arrive. You can find the code for the Arduino part here: https://github.com/shortstheory/quadcopter-experiments/blob/master/tristate.ino
12 Mar 2017
Recently, I’ve been trying to read more long form literature. It’s been an interesting experience to wane myself off of social media and that experience certainly deserves a post of its own soon. Until then, here are some of my thoughts on the books I finished reading this year:
The Checklist Manifesto - Atul Gawande
The title gives away the premise of the book - which is about the art of managing and creating efficient checklists. This may not exactly sound like a jaw-clenching thriller, but Atul Gawande manages to keep the pages turning with interesting anecdotes about maintaining checklists. A surgeon by profession, Gawande lists instances about how his carefully curated surgery checklist saved lives in the operating theater. As an aviation geek, I enjoyed the story of how creating a pre-flight checklist for the Boeing 299 (now known as the Boeing B-17) saved Boeing from the brink of bankruptcy in the fiercely competitive aviation market of the 1940s. There are narratives about the flip side too. For example, verbose checklists with too many steps for the user to follow, may do more harm than having no checklist at all.
For the most part, the book reads as a collection of anecdotes of varying quality. Pointers on creating a good checklist is scattered too thin throughout the book to be useful. That said, my opinion of this book is probably not the best one to go by as I finished the book over a period of months which broke the continuity of the book.
I’ve seen this book practically everywhere. I’ve heard praises sung about it at length from friends. I’ve even been listening to the Freakonomics radio podcast for quite a while. But, it was only last month I picked up this book. The cover makes bold claims such as “…Explores The Hidden Side of Everything” along with some terms of endearment from Malcolm Gladwell. Coming to the content, the book covers a range of bizarre topics such as ‘What do schoolteachers and sumo wrestlers have in common?’ and ‘Why Do Drug Dealers Still Live with Their Moms?’. Which is great, because the book makes convincing conclusions on these two topics. However, this cannot be said for some of the other chapters such as ‘What Makes a Perfect Parent?’ and how a child’s name is detrimental to their future prospects. I feel the book doesn’t spend enough time establishing how a certain cause leads to an effect in these chapters. The authors do include a pretty long list of citations for each fact at the end of the book (nearly as long as the content itself!), so their claims might be substantiated after all.
Russian Roulette - Anthony Horowitz
It took me forever to find this book. I used to be an avid fan of Alex Rider in school, but I lost track of the series in my dry spell of reading in 11th and 12th. Russian Roulette is something of a spin-off book, set somewhere in the middle of Stormbreaker in the Alex Rider timeline. The book is about all about Yassen, a key character in the Alex Rider series. Like most Anthony Horowitz books, it’s not a particularly well-written book. Rather, it keeps one engaged with a fast (but somewhat predictable) plot. I liked some of the parts from Yassen’s past in Russia, though the rest read like a rehash of Scorpia. Still, I finished this book much faster than I usually take to finish a book of its length.
Bird by Bird - Anne Lamott
In this book, Anne Lammot speaks about the joys and frustrations in the solitary world of a writer. From the outset, I loved this book. Lammot’s writing style is hilarious and is loaded with dry wit and sarcasm. As with The Checklist Manifesto, the book follows on tangents with no central plot to speak of. On my first read of Bird by Bird, I couldn’t help feeling that this book would read better as a blog than a novel. Though it of a much higher quality than a typical blog, the rant filled content would slide in perfectly for this format.
Lammot also peppers the book with some very quotable passages. I won’t spoil it here, but it’s a must read for anyone into the publishing industry or writing business.
05 Feb 2017
Now I finally have my very own domain name! The old Blogger site is still available if anyone wants to see it but for all future intents and purposes, this will be the place where new blog posts will be put up. While I could configure all traffic to the Blogger site to redirect to this one, the old site has grown on me so much over the years that I feel that it would be a shame to hide it that way.
I had to migrate from Blogger because it is more than evident that Blogger has been getting less love from Google than it deserves. Take for instance, the Blogger web-based post editor - a shining example of an undeveloped relic of Google’s products. Posts never autosaved or had any manual version control and a few misplaced keypresses could cause you to lose all your writing progress. Inserting images is a chore and putting up more complicated parts of text such as code blocks and sub-headings is even worse.
Of course as with most things in software development, there are workarounds for everything. The work-around which I had used for the last few blog posts was to write directly in Markdown and then convert it to HTML with pandoc to copy paste in the Blogger editor. Even so, this is clearly sub-optimal and I was spending more time wrestling with the Blogger editor to make my pages look good than I was spending writing actual content. The Blogger theme which I was using appealled to me as 13 year old (heck, it still appeals to me), but it was growing long in the tooth and only had basic support for responsive design and mobile devices. With all these things in mind, I started looking for a new home for my blog somewhere in the middle of last December on a break from college.
I first looked at WordPress. At first, it seemed like my search for a new blogging platform would stop here. Open-sourced, a great browser editor, a cohesive Android app, direct editing in Markdown, local installation for testing, and support for plugins made WordPress everything Blogger wasn’t. I loved the number of themes and customisation WordPress provided. But the love was short lived, and it ended when I started researching hosting options for WordPress. Most solutions required me to rent a web-server on a monthly basis and I had no idea what tier of server to get as my blog had only recently seen a huge surge in pageviews. Not to mention, the cost of maintaining a website with such a setup was by no means cheap. This is when I started asking to my geeky friends about how they maintained their own personal websites.
The talks were very helpful, it made me realise that a dynamic blogging solution such as WordPress was overkill for a humble blog of less than 30 published posts like this one. Having a static site made so much more sense. I could write directly in Markdown, in the text editor of my choice. Not to mention, it made the workflow of writing a blog post just like I wrote code,
git commit, and
git push straight to GitHub Pages. GitHub Pages, by the way, offers completely free hosting for static sites. This means that the only thing I would need to pay for would be the custom domain name, a nominal ₹700 ($10) a year, an amount half of what I would be paying for a month of paid hosting. Plus, it would let me get my hands dirty with a bit of web development, something which I had pushed back for a long time.
I decided to start this - what I knew was going to be painful task at about 11pm on a night I just knew I wouldn’t be able to sleep.
The first step was to convert all my Blogger posts to Markdown. There were some tools online but all of them messed up the conversion pretty badly. After some more digging, I ended up using Aaron Swartz’s html2text Python library which did a better job than other solutions in generating some useful Markdown. I still needed to edit every generated Markdown file by hand to make it something I would be happy with using on my site. I then had to export all the images I had on my Blogger site. This lead to a few more laborious hours of saving each image on the site by hand (Right Click -> View as Image -> Save As). It did cross my mind to automate everything with a script, but it was going to take more time to automate everything and check if the automation was working than it was to do the grunt work of pulling the images. With all the resources safely on my laptop and backed on my Dropbox, I took the next step of looking at static site generators to convert my lovingly handmade Markdown files to HTML.
GitHub Pages seemed to heavily advocate Jekyll so I went with it first. With some tinkering to get the Ruby dependencies installed and posts adapted for Jekyll with the Front Matter content, I managed to get a pretty presentable blog running on localhost:4000/ at 5am on that day. With a quick push to my github.io site, I decided to call it a night and slept off a sleep-deprived session of hacking.
The next few days I played with some more Jekyll themes and found that there were many things I didn’t like about it. For one thing, it was written in Ruby which I have no experience with. Themes didn’t look easy to work with and there was no native support for tags (there is a workaround for this, but due to my lack of Ruby-fu, it all looked terribly arcane to me). I then put the blog migration on the back burner for a while to work on projects at my college’s Automation & Robotics Club.
A few weeks later, I took a look at the blog project with some new perspective. I started by poking around for alternatives to Jekyll. There was one such alternative which ticked all my boxes - a static site generator called Pelican. As WordPress looked inherently superior to Blogger, Pelican looked inherently superior to Jekyll for what I wanted to do with it. For example, it had built-in support for tags, had a theming engine, supported Markdown and reStructuredText, and had several easy to install plugins. Above all, Pelican is written in Python which made it so much easier for me to mess around with it. There were some more modifications to make to the Markdown files (particularly with the post metadata), but it was so trivial that it didn’t pain one bit to modify all the files. Not too long after I settled on Pelican, I found a theme which made my blog look exactly how I wanted it to look. The Pure Single theme also has nifty support for custom sidebar images.
There was some initial trouble with setting a blog subfolder in the site and getting images to work on some auto-generated sites (such as the Tags and Categories pages). It later turned out that it was some problem with localhost/ not finding the paths correctly to the images and the site was totally fine when published to the GitHub Pages site. After only three days of using Pelican, I had something which I was willing to show off. The next step was much more straightforward for a change - registering a domain name. I looked into a few options such as GoDaddy, Hover, and Namecheap. Namecheap had positive reviews (unlike GoDaddy) and was the cheapest of the lot. The site configuration to serve pages from GitHub’s servers was not more than a 10 minute procedure, and I finally had the site you are reading this article on right now.
There will be a lot more changes coming up on this blog, some of them aesthetic and some functional. I’m also probably going to change the name of this blog sometime soon, to something which is more reflective of my current sensibilities.
19 Dec 2016
This semester has gone a little too fast for my liking. Though there were plenty of good memories and takeaways from this semester, it has been by far my most stressful time in this college yet. Between taking on a number of technical projects and trying to rescue a sharply falling SG due to a poor T1, there has been very little free time this sem. Above all, this sem has shown me just how limited of a resource brain power is. Thinking about solving problems wears me out much more easily than it used to.
That’s not to say what I’ve been doing is all drudgery. Some interesting projects that I’ve been working on:
GSoC 2016 Project - kio-stash
Yup, this project has been in the pipeline for months. While it (mostly) works on a clean install of KDE, it has some bugs with copying with mtp:/ device slaves and isn’t very well integrated with Dolphin yet. It is in my best interest to have this shipped with KDE Frameworks as soon as possible, so I’m looking into patching Dolphin with better, more specific action support for my project.
IEEE Signal Processing Cup 2017 - Real-Time Beat Tracking Challenge
The premise of this project doesn’t seem very complicated - a team just needs to submit an algorithm which can successfully detect the beats in a song in real-time on an embedded device. Heck, I had already implemented the same on an Arduino with a microphone for rather fancy lighting in my hostel room. Unfortunately, as I’ve learned the hard way, anything that looks simple is deceptive and this project is a perfect example. For one thing, a song by itself in the amplitude-time domain is nearly impossible to extract any data from, so it requires a (Fast) Fourier Transform to extract frequency bin amplitudes to get anything useful from it. Next, obtaining beat onset and detection requires generating a tempogram to find the BPM of a song to find the approximate beat positions. There are many interesting approaches and different algorithms to solve this problem. Fortunately, I’m not doing this alone and the seniors I’m working with have a better understanding of the more difficult mathematical components of this project, leaving me free to code in C++ and deploy it on actual hardware.
This project is a society on campus I started with threeprogrammers far more competent than myself. It all started in September this year when we had the two-day long Wikimedia Hackathon on campus. Though my contribution to Wikimedia was small, I got a lot out of the hackathon by meeting the people from Amrita University who came for organising the event. Listening to what Tony and Srijan had to say about how much working for foss@amrita had benefited them made me realise how much room for improvement there was in BPHC. As far as this society goes, things are still in their infancy, yet I feel there is just enough critical mass of talent and enthusiasm in open source development to push this society forward. We even have our own Constitution!
Making a personal website?
While Blogspot has been a faithful home for this blog for over half a decade, it is time to move to a modern solution for hosting. Plus setting up a website will teach me a bit about web development too, and I’ve needed an excuse to learn it for a while. Surely it can’t be all that hard, right?
While the slew of all these technical projects seemed like a good idea at the time, it has pushed me closer to burning out. It is the same feeling as what I felt the same time last year, albeit for different reasons.
Some of it has carried over from studying harder at the end of this sem than I had done so before. To my surprise, studying under pressure was actually more effective than it was without. In some ways, it was a throwback to studying for the number of exams nearly a couple of years ago. It has made me question whether any of my education was as much for the knowledge as it was for the desperation of avoiding failure. I still haven’t figured it out, but at least I feel like I’m not as adverse to putting in the hours for getting a decent CGPA as I had been last semester.
There have been compromises in other spheres of college life as well, investing so much time in studying or doing projects has made me cut down my participation in other clubs and I’ve quit nearly all the non-technical clubs I joined last year. There’s more than that too - I feel it’s much harder to be fully immersed in anything anymore.
Maybe it’s just something else which requires more soul searching.
29 Aug 2016
That’s it. After a combined total of 217 git commits, 6,202 lines of code added, and 4,167 lines of code deleted, GSoC 2016 is finally over.
These twelve weeks of programming have been a very enriching experience for me and making this project has taught me a lot about production quality software development. Little did I know that a small project I had put together in a 6 hour session of messing around with Qt would lead to something as big as this!
There have been many memorable moments throughout my coding period for the GSoC - such as the first time I got an ioslave to install correctly, to writing its “Hello World” equivalent, and getting a basic implementation of the project up and running by doing a series of dirty hacks with Dolphin’s code. There were also times when I was so frustrated with debugging for this project, that I wanted to do nothing but smash my laptop’s progressively failing display panel with the largest hammer I could find. The great mentorship from my GSoC mentor and the premise of the GSoC program itself kept me going. This also taught me an important lesson with regards to software development - no one starting out gets it right on their first try. It feels like after a long run of not quite getting the results I wanted, the GSoC is the thing which worked out for me as everything just fell into place.
There’s a technical digression here, which you can feel free to skip through if you don’t want to get into the details of the project.
Following up from the previous blog post, with the core features of the application complete, I had moved on to unit testing my project. For this project, unit testing involved writing test cases for each and every component of the application to find and fix any bugs. Despite the innocuous name, unit testing this project was a much bigger challenge than I expected. As for one thing, the ioslave in my project is merely a controller in the MVC system of the virtual Stash File System, the Dolphin file manager, and the KIO Slave itself. Besides, most of the ioslave’s functions have a void return type, so feeding the slave’s functions’ arguments to get an output for checking was not an option either.
This led me to use an approach, which my mentor aptly called “black box testing”.
In this approach, one writes unit tests testing for a specific action and then checking for whether the effects of the said action are as expected. In this case, the ioslave was tested by giving it a test file and then apply some of the ioslave’s functions such as copy, rename, delete, and stat. From there, through a bunch of QVERIFY calls is to check whether the ioslave has completed the operation successfully. Needless to say, this approach is far more convoluted to write unit tests for as it required checking each and every test file for its properties in every test case. Fortunately, the QTestLib API is pretty well documented so it wasn’t difficult to get started with writing unit tests. I also had a template of what a good test suite should look like thanks to David Faure’s excellent work on implementing automated unit testing for the Trash ioslave. With these two tools in hand, I started off with writing unit tests shortly before the second year of college started.
As expected, writing black box unit tests was a PITA in its own right. The first time I ran my unit test I came up with a dismal score of 6 unit tests passed out of the 17 I had written. This lead me to go back and check whether my unit tests were testing correctly at all. It turned out that I had made so many mistakes with writing the unit tests that an entire rewrite of the test suite wasn’t unwarranted.
With a rewrite of the test cases completed, I ran the test suite again. The results were a bit better - 13 out of the 17 test cases passed, but 4 failed test cases - enough reason for the project to be unshippable. Looking into the issue a bit deeper, I found out that all the D-Bus calls to my ioslave for copy and move operations were not working correctly! Given that I had spent so much time on making sure the ioslave was robust enough, this was a mixed surprise. Finally, after a week of rewriting and to an extent, refactoring the rename and copy functions of the ioslave, I got the best terminal output I ever wanted from this project.
Definitely the highest point of the GSoC for me. From there on out, it was a matter of putting the code on the slow burner for cleaning up any leftover debug statements and for writing documentation for other obscure sections. With a net total of nearly 2000 lines of code, it far surpasses any other project I’ve done in terms of size and quality of code written.
At some points in the project, I felt that the stipend was far too generous, for many people working on KDE voluntarily produce projects much larger thann mine. In the end, I feel the best way to repay the generosity is to continue with open source development - just as the GSoC intended. Prior to the GSoC, open source was simply an interesting concept to me, but contributing a couple of thousands of lines of code to an open source codebase has made me realise just how powerful open source is. There were no restrictions on when I had to work (usually my productivity was at its peak at outlandish late night hours), on the machine I used for coding (a trusty IdeaPad, replaced with a much nicer ThinkPad), or on the place where I felt most comfortable coding from (a toss up between my much used study table or the living room). In many ways, working from home was probably the best environment I could ask for when it came to working on this project. Hacking on an open source project gave me a sense of gratification solving a problem in competitive programming never could have.
The Google Summer of Code may be over, but my journey with open source development has just begun. Here’s to even bigger and better projects in the future!