blog.zephod.com

Lego/Raspberry Pi Project, a set on Flickr.I kept a photo diary...

Wed, 19 Dec 2012 11:58:47 +0000

Lego/Raspberry Pi Project, a set on Flickr.

I kept a photo diary during the construction of the Lego/Xbox/Raspberry Pi project.

Build your own RC Car using Lego, an Xbox Controller and a Raspberry Pi

Mon, 03 Dec 2012 17:41:00 +0000

When I was a child I always wanted a remote control car. I never got one — usually my parents convinced me that the novelty would wear off in an hour, and that it wouldn’t be worth the expense. (They were probably right). I usually opted for a safer videogame-based option at Christmas time, and on my birthday I’d ask for Lego.

The Raspberry Pi — a tiny, low-power computer that fits in the palm of your hand — provided the inspiration for a project to finally put that right. Using bricks from my childhood Lego collection, and a wireless Xbox 360 controller from my teenage years, at long last I could hack together a remote control car! How hard could it be?

Not exactly simple

I’m a programmer by trade, but any knowledge of electronics learned during University apparently died at graduation. The necessary software configuration and ~200 lines of Python coding took very little time. The fairly simple circuitry took a long time, and caught fire on several occasions. Your mileage may vary.

There have been similar attempts to control RC cars from an Xbox controller:

Dan from Chicago put a Netduino controller on his car which connects via his laptop.
Keegan at jedibowler.com used Netduino on the car as well, and has a PC on the floor issuing commands.

This project takes it a step further because the Raspberry Pi is a complete computer, so all the hardware lives on the car. There is no need for a laptop issuing commands over Wifi.

Building the Chassis

The design of the chassis aims to be as simple and robust as possible.

No, really, robustness is key. Some observations:

Use Technic clip-joints and axles through everything.
Using Lego studs to join any two parts of the chassis was a waste of effort. Shearing stress pulls them apart easily.
Servos and motors would rather pull the Lego apart than turn the wheels. They need to be very firmly held in place.

Two motors are required:

The steering is controlled via a high torque 180 degree Towerpro servo from AdaFruit. A Lego gear is attached with superglue.
The rear axle is attached to a Lego Power Functions XL-Motor.

Originally I used an old 9V Technic motor, but it wasn’t strong enough to turn the rear axle. There’s a fascinating analysis of Lego motors on philohome.com which suggests that the newer motor has around 40 times more torque. I purchased one online for just $10.

Electronics and Soldering

The motors are controlled via PWM (Pulse Width Modulation) using the excellent Adafruit 16-channel PWM driver.

The tutorials on Adafruit’s website are truly exemplary, and in less than an hour I had the Raspberry Pi controlling motor speed and servo rotation.

The above shot demonstrates circuitry mounted onto the chassis.

Adafruit’s Cobbler Breakout Kit is used to connect the Raspberry Pi to a prototyping plug-board.
The i2c connectors (SDA0 and SCL0, 3v3 and GND) are then connected to the PWM driver.
GPIO pins 21 and 17 are connected to a cheap H Bridge chip, specifically the TB6612FNG from pololu.com. This will drive the main motor forwards or backwards.
Output 0 on the PWM driver runs into the H Bridge. This controls the main motor’s speed.
Output 1 on the PWM driver runs to the steering servo.

Software

Finally, the Raspberry Pi must be loaded with some software which translates Xbox controller buttons into motor speed, motor direction, and steering rotation.

The Xbox 360 Wireless Gaming Receiver for Windows is required to connect the controller. Microsoft use a custom wireless protocol, so a regular Bluetooth modem will not work. This bulky device consumes quite a lot of power, but at least it is relatively inexpensive.

“For Windows” should not be taken too seriously. Ingo Ruhnke’s excellent Xbox/Xbox 360 USB Gamepad Driver for Linux is a perfect solution for hooking up the wireless USB device, and it installs in seconds on a Raspberry Pi running Ubuntu.

At last, I can get back to programming! I hacked together a couple of Python modules to:

Run xboxdrv as a subprocess, which connects to the controller.
Read from the stdout pipe as it constantly prints out the controller state.
Watch for changes in LT (left trigger), RT (right trigger) and X1 (left thumbstick).
Update the outputs as appropriate through Adafruit’s libraries.

For example, to control the steering:

# Left thumbstick controls the steering
if event.key=='X1':
    steer = int( servoMid + (servoWidth*-event.value)/32768 )
    pwm.setPWM(1, 0, steer)

The complete software kit is available in a repository on Github.

Completion

Here’s my DIY-RC-Car in action. Enjoy!

Tom Rees is a freelance hacker, developer, and code poet. You can follow him on Twitter if you like this sort of thing.

Hacking Scrabble

Fri, 27 Jul 2012 10:41:00 +0100

For the lazy: Not interested in all this data wrangling? Skip straight to these summary graphics to improve your Scrabble game:

Introduction

A while ago I was blogging about a little research project I called Hacking Scrabble, which received a lot of interest and was published on LifeHacker. I was trying to get better at the game by using the kind of memory tricks we learned in school:

“My Very Easy Method Just Speeds Up Naming Planets”

I’m a programmer by trade so inevitably I wrote Mnemonic Finder, a Python program to quickly search the dictionary looking for these kinds of phrases. The idea was to memorise all two-letter words, which I’m targeting because:

They are vital. Games become stagnant when the board is full up and nobody can play in the tight corners without creating a string of two-letter words.
There aren’t actually that many of them: 124 in the international rules, 101 in the US rules.

The Solution (First Attempt)

My first attempt at writing this column arrived at a technique which sort of worked.

Every 2 letter word is a prefix and a suffix.
- eg. “at” has a prefix “a” and a suffix “t”
So every prefix can have a few different suffixes.
- The prefix a can have suffix t but not suffix q, because aq is not a word.
For each letter, list allits possible suffixes.
- a can have any of: abdeghilmnrstwxy
Use Mnemonic Finder to look for a phrase which contains all those letters but no other letters.
- “birthdays mangle wax” will help you remember all the letters that can go after a.

Why Doesn’t It Work?

First, it’s not that great. I still need to memorise twenty six phrases and know which letter they each apply to.

Second, the suffixes of u are ghmnprst. Can you see the problem? There are no vowels! ua, ue, ui, uo, and uu are not words. So Mnemonic Finder can’t find a single thing in the dictionary. Maybe you can learn it by saying it out loud: “Gahumnapurst”. Maybe not.

In fact, out of twenty six attempts, I only succeeded in finding about ten useful phrases. Okay, it’s better than learning over a hundred words, but we can do better still with a little lateral thinking.

Restructuring The Problem

Many astute readers emailed me to point out that nearly all two-letter words contain a vowel. Instead of memorising the suffixes of every letter, we could memorise the suffixes and prefixes of every vowel.

               << prefixes    vowel   suffixes >>
           ------------------------------------------
            ztnlkjhdbfypmae     a     aeibdghlmnrstwxy
             ywtrpnmhfdbaeo     e     aedfhlmnrstx
             xtsqpmlhkdgboa     i     odfnst
          zywtsmlkjhbdgnpio     o     eioubdfhmnprswxy
                     gmnxyo     u     ghmnprst

This is a great idea: We only have to memorise ten phrases… plus a few words containing no vowels.
However, the computer still has trouble finding words in the dictionary… “ghmnprst” is still up there.
There’s a bit of redundant information in there. o is both a prefix and a suffix of itself because oo is a word.

We’re on the right path. A final step is needed to make this work.

Just The Consonants, Please

The problem is that it’s hard to find mnemonics if you don’t have many vowels. I experimented with a tweak to Mnemonic Finder allowing it to use all the vowels it likes, and once again gave it ghmnprst.

    "prognathism"
    "graph mints"
    "hemp string"
    "nightmare gasp"
    ...

Thousands of phrases tumbled out! This is the key to the solution: Let’s seperate the vowels from the consonants. Here’s the table again with only consonants in the left and right columns.

                << prefixes   vowel   suffixes >>
            ------------------------------------------
              ztnlkjhdbfypm     a     bdghlmnrstwxy
                ywtrpnmhfdb     e     dfhlmnrstx
               xtsqpmlhkdgb     i     dfnst
            zywtsmlkjhbdgnp     o     bdfhmnprswxy
                      gmnxy     u     ghmnprst

The words we had to remove were the double vowels. A shame that they didn’t fit, but they won’t be too difficult to memorise: There’s only nine of them (or just five for players in the USA).

    aa ae ai ee ea io oo ou oe

Finally, I’ve avoided telling you about the double consonants. They never fitted onto that table anyway, and must also be memorised the hard way:

    sh ch st mm hm my ky fy by ny

Through the Mnemonic Machine

Using Mnemonic Finder in “free vowels mode” produces hundreds (usually thousands) of dictionary words for each entry in the table. At last, we can replace ugly blocks such as ghmnprst with meaningful, memorable phrases like hemp string.

                    << prefixes   vowel   suffixes >>
      ------------------------------------------------------------
  the band played folk jazz metal   a   birthdays mangle wax
             thumbprint fade away   e   tax his farmland
     exquisite godlike humble pie   i   infested
      mask playing with jazz band   o   six dumb hippy windsurfers
                         onyx gem   u   hemp string

You can generate your own if these aren’t to your taste. I prefer phrases with a powerful (if absurd) mental image. I’ve logged my development in a public gist for both the International version and the US/Canada version.

It is important to remember that we are just memorising consonants. The suffixes of i are “infested”, meaning n, f, s, t, or d. Ignore the vowels; ie is not a word.

Once you understand and memorise that table, you know one hundred and four words! The remaining twenty I leave to you. Did you think it was all going to be that easy?

    aa ae ai ee ea io oo ou oe 
    sh ch st mm hm my ky fy by ny

In A Real Life Battle Situation...

So you want to play a two-letter-word, but you’re not sure it exists?

If it’s a double consonant, you should know it by heart.
If it’s a double vowel, you should also know it by heart!
If it starts with a vowel, think of the right hand column.
- Eg. The first letter is e, the second letter must be in tax his farmland.
If it ends with a vowel, think of the left hand column.
- Eg. if the last letter is e, the first letter must be in thumbprint fade away.

Fortunately, I have turned all of this into a beautiful infographic that you can print out and treasure forever.

For players using the North American rules (US, Canada, and Thailand):

For players using the International rules (everywhere except the US, Canada and Thailand):

I am Tom Rees, a freelance hacker and engineer. I write lots of code for the Open Knowledge Foundation.

You should follow me on Twitter.

The International version of the Scrabble memory technique I...

Tue, 24 Jul 2012 23:04:22 +0100

The International version of the Scrabble memory technique I have been researching. (Valid everywhere except the US, Canada and Thailand).

Source data here: https://gist.github.com/3141231

The US version of the Scrabble memory technique I have been...

Tue, 24 Jul 2012 22:33:00 +0100

The US version of the Scrabble memory technique I have been researching. (Valid in the US, Canada and Thailand).

Source data here: https://gist.github.com/3141380

Hacking Scrabble (part 1)

Tue, 13 Dec 2011 15:11:00 +0000

Edit: See this article for the final result of this project!

This post isn’t really about Scrabble. It’s about taking a load of ugly data and hacking around with some scripts to refine it into something I can commit to memory. Then its about Scrabble. Winning at Scrabble.

The Problem

Knowing exactly which two-letter words exist gives a player a sizeable advantage because it opens up tight corners of the game board and allows them to run two words side-by-side. Sadly most of us only know a handful, and beyond that we’re guessing. Then we’re arguing. And then (inevitably) looking on Google for a list of valid words.

Personally I’m all for having the 2-letter-word-list on display during play to help us out of tight spots and to keep things fair. But I’ve had plenty of opponents object to this, saying that if I really want to play like that I should memorize the list. How hard can it be?

aa ab ad ae ag ah ai al am an ar as at aw ax ay ba be bi bo by ch da de di do ea ed ee ef eh el em en er es et ex fa fe fy gi go gu ha he hi hm ho id if in io is it ja jo ka ki ko ky la li lo ma me mi mm mo mu my na ne no nu ny ob od oe of oh oi om on oo op or os ou ow ox oy pa pe pi po qi re sh si so st ta te ti to ug uh um un up ur us ut we wo xi xu ya ye yo yu za zo

(source: SOWPODS, used for British tournament play)

That’s 124 words! More than a hundred arbitrary letter-pairs, with no connecting pattern and most of which I don’t recognise. Committing that to memory is better left to sauvants who can happily recall the order of a shuffled deck of cards.

I had already given up with the problem, and I was writing the list out on paper to be kept in the game box. To make it easier to fit on a page, I tried to compress the information, producing a table like this:

a: abdeghilmnrstwxy   j: ao                 s: hiot
b: aeioy              k: aioy               t: aeio
c: h                  l: aio                u: ghmnprst
d: aeio               m: aeimouy            w: eo
e: adefhlmnrstx       n: aeouy              x: iu
f: aey                o: bdefhimnoprsuwxy   y: aeou
g: iou                p: aeio               z: ao
h: aeimo              q: i                 
i: dfnost             r: e

Suddenly it’s not so terrifying, is it? The data has some structure. Every prefix has a number of possible suffixes.

b can be followed by a, e, i, o, or y.
But c can only be followed by h.

It’s still a lot to commit to memory. Too much: a and o each have sixteen possible suffixes!

The Mnemonic Method

I still believe it can be done! The key is in finding an appropriate mnemonic to help us remember arbitrary sets of letters.

So let’s begin with all the two-letter-words starting with a. If I can memorize the set of letters ‘abdeghilmnrstwxy’ then I can mentally verify any word that starts with a by asking, ‘is the suffix in that set?’

The Mnemonic Method: Anagrams

Your first thought might be to search for anagrams of the string. Mine was too. You’re out of luck - just try asking Wolfram Alpha for anagrams of abdefhilmnrstwxy. The problem is simply that there are too few vowels available to build good words.

The Mnemonic Method: Word Subsets

Try asking Wolfram Alpha for word subsets of abdeghilmnrstwxy. Now we’re on to something - it returns 4243 results.

a | Abe | abed | Abel | abet | abets | abide | abides | able | abler | ables | ablest | ably | Abner | abridge | abridges | abridgment | abridgments | abs | ABS | abseil | absent | absently | absinthe | … (total: 4243)

These are words containing only those letters. If we can memorize enough of those words to cover all sixteen letters (‘axe labs abridgements…’) then we can mentally verify any two-letter-word starting with a by asking ‘is the suffix in any of those words?’.

A couple of observations at this point:

Hand-picking the right words from 4243 results is painful.
Words like beer are missing because they re-use the letter e.

For the purpose of this memory-check, we don’t care if a letter appears more than once. To memorize the letters elh, we might remember it as ‘hell’ - but this method will not suggest that word.

The Solution

Wolfram Alpha’s algorithm is based around the following operation:

Find every word in the list of letters, and check it is in the dictionary.

We can find many more words by reversing the process:

Find every word in the dictionary, and check only uses letters in the list.

I went on to hack together this Python script which does exactly this. It allows us to control specific things like minimum and maximum word length and allows reuse of letters, so finds many more words than Wolfram Alpha can.

Armed with this huge number of words, we can programatically look for groups of them which cover all the letters, and save them as mnemonic devices.

The 100-line algorithm runs as follows:

Load up a good British word list (obtained from anywhere on the internet).
Filter the list to remove words which contain the wrong letters. Using the letters ‘abdeghilmnrstwxy’, beer remains in the list while vodka is removed.
Find the word which uses as many different letters as possible. Eg ‘nearsightedly’ which uses twelve of the letters. This is the start of the mnemonic.
Keep finding words which use up the remaining letters.
Loop until all letters are covered. Eg. “nearsightedly wax balm”.

Demonstration

This is it! Here I am, searching for mnemonics to help me remember the two-letter-words starting with a.

$ python mnemonicfinder.py abdeghilmnrstwxy --minlen 3 --maxlen 9
==> MnemonicFinder v1.0
--> Loading dictionary...
--> Search string: abdeghilmnrstwxy
--> Generating...
--> Mnemonic: ['bearishly', 'midwing', 'tax']
--> Mnemonic: ['berthings', 'madly', 'wax']
--> Mnemonic: ['betraying', 'mildews', 'hex']
--> Mnemonic: ['birthdays', 'mangle', 'wax']
    (output clipped)

Okay, some are better than others. By scanning the output for a sentence or phrase which I can attach to a mental image, I can make it memorable: “birthdays mangle wax” means candles on a birthday cake. Is ap a word? No, because there’s no p in that phrase.

Hooray! I am one twenty-sixth of the way towards memorising all of the two-letter-words permitted in Scrabble. Have a look at the files on GitHub if you want to try out the script for yourself. In Part 2, I’ll develop 26 mnemonics and work on a technique for memorising them.