A really good friend of mine lent me this book, which I recently managed to devour in just about 2 days:
<photo from goodreads.com>
Written by a guy that did an undergraduate degree in computer science and philosophy, and an MFA in poetry, the book was ultimately a fascinating exploration of how an AI compares to a human, and what place AI might (or should) eventually have in society.
With topics ranging from Information Entropy, the Turing Test, the Loebner Prize, Deep Blue vs Gary Kasparov, rendering algorithms and Game Theory, the book presents itself almost like a steam of consciousness, as the author explores the effect of AI and the advent of reasoning in our society.
I came away from reading this book learning just enough about all these topics to whet the appetite, but not really enough to shed light on any truly difficult questions…
The lorry with our goodies!
Haha it came back in pieces.
Taking up space in Fab Lab 2
Our go karts are finally back from MIT!! Hilarity ensued when we found various nuts and bolts dropping out of the crates and we attempted to figure out where everything came from.
Charles has kindly shipped all his spare parts from the 2.00summer go kart course so we can continue to work on our EVs without having to order and ship extra stuff to SUTD.
It was a mild let down that our go karts arrived after both the Learning Carnival and the visit by the Solar Car team from Eindhoven – Learning Carnival was something that was organized from 16th-18th of Oct so that we could share with each other what we’ve all done over the summer. For us, a large part of it was the go kart experience that we had in MIT, and so we decided to try to build another go kart in < 4 weeks, so that we had something to show everyone else.
To me, Go Karts and Homemade-Motorized-Things-That-Move are important to everyone (especially engineers) not only because they’re hilarious and utter fun, but because it gives everyone an opportunity to build.
Now that we have our go karts, parts and supplies, and more importantly go-kart knowledge, an annual SUTD Grand Prix might actually become reality. Muahaha.
And so the emergency ev is done! We’ve managed to build it cannablizing parts from another go kart-ish thing inspired by the chibikart blog, but at least it’s done!
Nishard on the Emergency EV
Important things to note for the future:
- RS Components is a wonderful site as they have a warehouse in SIngapore, and they do next day delivery.
- Always check and double confirm the order that you’ve made both through your own screenshot and through the receipt they send through the email – RS Components’ website bugged out and resulted in me buying 3 bags of M12 nuts 3 times because they always sent me M8 instead and I assumed I clicked wrongly.
- McMaster has all the CAD files you could ever need.
- Waterjet can’t take files that are too complicated. Always let the auto pathing create the path and keep tweaking the file and configurations until it works – if you manually draw the paths there’s a chance that the waterjet will bug out, and in the worst case, destroy your material.
- Always buy spares of almost everything.
Now that the Emergency EV is over, I guess I could finally re-embark upon the mini amigo bot. Oh my.
We found out last week that our go-karts were not going to be back in SUTD on time for the 16th-18th Oct Learning Celebration Carnival.
We have thus decided to
magically make a go-kart appear make a working go kart in the short span of about 20 days.
Procurement so far has been terrible partly due to the lack of McMaster and Andy Mark and tncscooters and monsterscooterparts and all those other wonderful sites Americans
take for granted get to enjoy. Over here in Singapore, the two largest online vendors that deal with hardware in general are Misumi and RS Components.
On the plus side, we have our own waterjet, but we’ve got to convert all our cad to fit to the metric system – 80/20 rails are now 25mm thick (instead of 1 inch) and secured with M6 screws instead of quarter inch bolts…
By the time this post was written, most of the ordering has already been done. What remains to be seen is if we can actually get something moving…
Arron just told me about deal extreme, and it’s a great site for cheap
china electronics. One of the largest problems with the previous Arduino minime setup was the voltage regulator, which kept burning out – you can find a cheap buck converter (a much better piece of electronics which accomplishes the same function) on there for about 3.90, which is about as cheap as it gets.
With this, I could get a 3s or 12V battery pack and use the buck converter to get the voltage down to 5V so the Arduino could accept it safely (but actually the arduino can accept up to 12V so this isn’t needed, and if it must it can take up to 20V) but in the interest of preserving the Arduino and having a relatively constant voltage supply this is probably the best way to go about it. 12V is also the recommended voltage for most dc motors, so a 12V pack is generally a good idea.
You could also get interesting things like a dc motor controller, which can take up to 800mA, perfect for two low power Pololu micrometal gear motors rated for 0.36A each.
And whaddyaknow shipping to Singapore is free. Win.
We have to rush crazily before the academic term begins to get our stuff up, and we don’t want to sit on this project for too long.
Endpoint: A small robot that can move around at a rough speed of about 10cm/s and be able to sense it’s surroundings with two sonars that can be swivelled about. This robot must be able to interface through a serial port with a mac/pc/linuxbox running Python code. Sample python and arduino code must exist to show that this robot can run simple algorithms such as AngleProp and DelayProp.
From now till 16 Sept (start of term): Fabrication and Cadding
Term 4 Week 1-4: Coding and Testing
Term 4 Week 5-6: Testing and Preparation for presentation
Term 4 Recess Week (week 7): Presentation
Term 7: Next batch of 3.007 happens, and if this works out we’ll be providing support.
In term 2 we were tasked to build a “minime” for 3.007, which was a small vehicle powered by a rechargeable battery pack and controlled by an Arduino.
The great minime
Largely to teach us fabrication methods, we spent the week using Solidworks to CAD the minime and then sawing and bending acrylic to make the chassis. We then connected a bunch of wires to a breadboard and attached the servo motors and we had our very own robotics car controlled by a photoresistor (turn left if light, turn right if no light etc.)
And then in term 3, we had the AmigoBot for use in Digital World, where we experimented with Signals and Systems and used algorithms to turn sensor input into movement output.
But the AmigoBot was not ideal for a number of reasons:
- They spoilt easily, despite having a polycarbonate shell etc. This also included the port for the wire, which somehow had really bad connectivity.
- It was heavy, which meant that it spoilt easier. This also meant that algorithms were inherently significantly off due to a lack of ability to account for inertia.
- They were not repairable in house, so we always had to get someone to repair them.
- They cost roughly $3000.
And so the miniamigobot was born – why pay $3000 for something you can pay a little more and get each student to make? While it is an added long term cost for the school to keep buying parts and giving them to students, I believe that the learning opportunity from a project like this can be well worth it.
All we needed was two sonars, two servo motors to swivel those sonars around, and voila we’ll be done… or so we thought. After some research and collection of scrap electronics lying around we realized a few things:
- Sonars come in many shapes and sizes. And costs.
- Getting accurate feedback from each wheel is not going to be easy – there are two ways to do this, with a stepper motor or with a dc motor that comes with an encoder.
- But using a dc motor would require a different arduino, more specifically, a dfrduino romeo.
- Getting python (the DigiWorld course is run in python) to talk to the Arduino over serial might not be as easy as we thought.
And so this led to the new to do list:
- Test different sonars on different surfaces and angles to find the most cost effective solution (we’re all about $$ here)
- Test various feedback mechanisms
- Research on (dc controller + arduino) vs (dfrduino romeo) vs (servo motor + hacked on feedback mechanism)
- Get python and arduino to talk to each other, prove that AngleProp and DelayProp both work.