Cueva secreta de Malte

I'm busy....

Wed, 16 Nov 2011 18:12:22 +0100

Right now I’m pretty busy writing a report about our specialization projects, that i will hand in on friday.

I have been working on the AI for the sentry gun, including some basic prediction of movement, but I still have a few hours ahead of me working on filtering motion rectangles, creating new targets and applying new motion to existing targets.

I just posted a picture of the disassembled airsoft rifle that is going to be mounted on the sentry. I hope i will be able to get it mounted before the 25th, where we will be presenting our projects for our teachers.

I most likely will not be able to finish the project before handing in the report, but i should be able to finish it before the presentations.

This is a aluminum gearbox, with a bunch of gears, some wires, a...

Wed, 16 Nov 2011 18:05:00 +0100

This is a aluminum gearbox, with a bunch of gears, some wires, a relay, a motor, and a laser sight.

The spring, which is the part deciding the muzzle velocity of the gun, is a tad too strong. Currently it creates a lot of vibrations, but that should decrease by hopefully a reasonable amount when I open it up and swap it with another.

This is what the XNA game i’m working on looks...

Fri, 28 Oct 2011 16:21:12 +0200

This is what the XNA game i’m working on looks like.
It’s a single player top down tanks game.

Currently I am working on making the tank move correctly.
The way it should move is that it’s given a direction, from the arrow keys, then it should rotate to that direction before starting to move.

The difficult thing about that is that it shouldn’t have a front or a back, like if i press up, it will move upwards, and if i press down, it will move down, but if i press left, it will rotate 90 degree and then move left.

I have been trying to make this game a few times before, but this tank rotation thingy is always the thing messing with my head, I am really counting on figuring it out this time though :)

One of the other things i’m trying to figure out, is how to mix 2D and 3D.
The dots between the two tanks on the image is the tracks from the green tank. Beside not being rotated correctly, they are on top of everything else.
Currently i’m thinking about placing a quad / plane in the 3D world, and then draw the texture on that instead. A quad / plane is a rectangle with no dimensions, as i only need one side, the one on which to draw the texture.

If there is one thing i find annoying and confusing in XNA / game development, it’s the different units and ways of expressing angles etc. Spent a good deal of time yeaterday trying to work out the tank rotation, only to find out in the end that the angle which i was using was not 0-360, instead it was -180 to 180 - much like radians. At that point i had deleted all the code since it wasn’t working…. But maybe it was?

Damn angles.

Next is collision detection, bullets, better lighting, explosions etc.

This is how my mandatory android application ended up...

Fri, 28 Oct 2011 15:34:00 +0200

This is how my mandatory android application ended up looking.

It is implemented using an AsyncTask that handles retrieveing the xml weather data from the unofficial google weather api, and displaying the updated information in the UI.
A nice progressbar is shown when the app is updating.

The code was pretty straight forward, i really like android dev, except the whole “java-deal” ;) I spent quite a bit of time working on the xml layout, it just wouldn’t seem to do as i wanted it to. In the end i discovered that most of the xml i wrote was correct, the issue was that it didn’t want to display it correct. That is untill i set the properties using code instead of xml.

Then it worked as a charm :)

Quick summary of android, xna and sentry gun stuff

Thu, 13 Oct 2011 21:55:49 +0200

Android
We currently have a small project going, a weather app fetching xml data on some webservice.

I decided to use a SAX parser, since it’s pretty damn quick and simple at the same time. Basicly you write a class that’s custom made for some specific XML scheme.

I made a model class that contains fields for all the info i get back from the webservice.
And then i used that model class to store the info that i got from the xml response.

Next up is some more playing around with Intents, activites and the styling.

XNA
I decided to scrap the 2D stuff, and go with something 3D instead.
I am creating a “Wii Tanks” clone, which is a top down tank game.

I want to replicate the feeling and the graphics of the game, so i spent quite a bit of time figuring out how to do that, and our teacher came up with the solution, which was using ortographic projection, together with rotating the camera around 30 degree. That nailed it pretty good.

After getting annoyed at all the 2D game engine classes i created a long time ago, I created a new XNA project and moved only the most important stuff. Hell, i made the other crap before, so I can do it again.

I really hope i will make it far enough to start working on the AI in the enemy tanks, the wii game has some pretty damn cool enemies. Which i by the way will never be able to replicate fully or even just a little. But I’m looking forward to trying :)

Sentry
I had a quick meeting with my supervisor yesterday, and it turns out that he knows a ton about cameras, which is pretty cool, since my project relies pretty heavily on a good framerate etc. After speaking to him i got quite a few new ideas.

I spent the last few days working on the architecture of my application. My delete button was red-hot, I have a bad case of “premature optimization”, which is exactly what it sounds like. I keep thinking that “this should take this and that into consideration, and what if.. bla bla.”. Sometimes i know when to stop, but other times i just keep going, until it’s too big and complicated for me to keep it in my head.

Thank god god made the delete button!

Currently I’m working on a small polling thread that updates info about the servo positions and targets in my application.

And after that i will start looking into the main “brain” of the sentry.
I’m looking to make it as intelligent as possible, but currently my mind is in “simple” mode, which means i can (hopefully) get shit done :)

Oh, and I’m really glad that i signed up for the Android sessions, instead of creating my own list of resources and studying at home. Once or twice *cough* I have had trouble getting motivated when at home.

Game development: xna and unity introduction

Wed, 28 Sep 2011 19:13:19 +0200

We got a quick introduction to XNS and Unity. In both engines/frameworks we created a simple 3D scene.

After the introductions was over, we sat down in groups and started playing around with xna and unity.

We decided to try and make a basic platform game with pixel based colission detection.

After a few hours we got a small very simple prototype done :)

Game Development Recap

Mon, 26 Sep 2011 14:46:24 +0200

We have spent some time figuring out what sort of game we want to develop, while doing so we used a few different techniques to help us sort of decipher what makes a good game.

That was done by each student listing three games they really like playing, and then trying to figure out why they play those games.
Developing the actual game is hard, but developing the game concept, the reason for people to play the game is REALLY hard.

Half through the day we had a note with three game ideas, and now we had to add some details to the game, like why, how, where and by who. Which i had a really hard time doing, because two of my ideas was missing the basic game concept, i still havent decided which one i want to keep working on, but i will probably end up working on two of the ideas.

The ideas i ended up with is

Tanks
“Build something”
Obstacle course (platformer)

I will probably ditch the obstacle course one, and then try to see if i can work up a concept for the “build something” idea.
I will use XNA for development, but currently i haven’t really figured out if i want to do a desktop game for windows, or a mobile game for WP7.

The programming is pretty much the same, so i should be able to change my mind a few times before deciding final.

Specialization Courses

Mon, 26 Sep 2011 14:28:00 +0200

I changed my mind. Instead of only attending the game development course, i also joined the android course, so i don’t have to spend too much time at home, which tends to kill all my motivation.

I also just really wanted to learn some more about android development. I mean, it would be dumb not to take the oppertunity to learn it. Just look at the current android marketshare, it’s HUGE!

Which means that i’m currently attending two courses, game and android development.

This blog will from now on also function as a diary of what i have been doing and what i have been learning in those courses :)

Motion capture performance and vibrations in rig

Fri, 16 Sep 2011 10:35:25 +0200

Motion capture

I tested the algorithms with the PS3 Eye camera, but unfortunately the algorithms are pretty slow, and it is still missing a lot of calculations, though i think i may try to do implement that using some kind of threading, perhaps producer consumer.

I don’t remember the data, but at first with some very basic calculations it was pretty good, like 10 ms per frame, but as soon as i plugged one of the AForge background motion detectors on it, the processing time per frame rose to what should be around 34 frames per second,which is still good, but i was hoping for a bit more.

I plan on spending some time on making another background detector, a more lightweight simplier one. But the one used right now got some really awesome code, so i guess my only option is to simplify and remove features..

I plan on trying out a more powerful laptop or even a ordinary tower, and i’m pretty excited to see how much it will improve.

Vibrations / Rig

A thing that has been on my mind for a while is vibrations…
The rifle that’s going to be mounted on top of the rig will produce some vibrations, the rig should handle it fine, but the camera and the algorithm will probably not be as forgiving. Motion is implemented as pixels in the image, changing color, so if the camera is shaking, either when firering the “dispenser” or by operating the servos, it will detect the whole frame as motion.

I should be able to (partially) solve that problem by somehow reducing the vibrations transferred from the rifle, by removeing weight and mounting using some vibration absorbing material. And in the algorithm i could make it “take a break” if every pixel changes.
But at the same time i don’t want to paralyze it when ever vibrations happen.

The last thing i can think of is mounting the camera on a seperate rig or something, But that makes calibrating the sentry a pain.

Algorithm wise, i may try to fiddle a bit with C++ since it seems QUITE a lot faster :)

I will be attending the Android course too, considering the current android market share :)
I have some trouble motivating myself sometimes, so i fear that i may just get right about nothing done if i spend too much time at home.

Oh yeah, i almost forgot.
The servo controller has a nice feature called acceleration, that should greately increase the lifespan of the plastic gears and servos. While also decreasing the vibrations and shaking when operating the servos.

Currently i have a kinda old 6 volt battery connected, but the servos i use can take as much as 7.2 volt, and i’m pretty sure that without the accelleration feature, the servos would smash the plastic gears in the first second of testing it :)

After all, combining the forces of those two bad boys would almost be enough to carry me.. (75 kg)

The 50-60 spelling mistakes are inserted on purpose, i dare you to find them (and tell me where they are)

The django book 2.0 on my kindle :) (Taken with picplz.)

Mon, 22 Aug 2011 15:22:08 +0200

The django book 2.0 on my kindle :) (Taken with picplz.)

Ja tak! (Taken with picplz.)

Thu, 18 Aug 2011 14:28:01 +0200

Ja tak! (Taken with picplz.)

Motion Detection Considerations Part 1

Wed, 17 Aug 2011 09:21:00 +0200

This is the first of a small line of posts about my thoughts and considerations building a autonomous sentry gun.

Link to initial post about the project

The project mainly consists of two important parts. The servo controller and the video feed/motion detection. In other words you could say input and output. This post is to clarify some of my considerations concerning both the video feed and the motion detection.

I have decided to use the PS3 eye camera, because it has the ability to capture images roughly twice as fast as a ordinary camera, at 640x480 resolution and like 4 times as fast at 320x240 or something. 640x480 at 60 GPS is my goal. Which is a great resolutions for this project. Because the smaller the images, the faster processing through the frames will be.

Camera bottleneck is then eliminated, but I’m almost certain that the 60 FPS camera will drown my computer in frames. Processing will just take too long for it to go through 60 frames a sec. At 60 FPS each frame has to be processed in less than 16.6 ms, which is not alot of time.

That will either require me getting a faster computer, or to use some optimized crazy fast algorithm.

No actual benchmark has been performed yet, so I really can’t wait to try it and get some data to look at.

I have decided to use a open source machine vision library called AForge to help me with the actual processing of the frames. It is either that or another one called OpenCV. But there are some nice algorithms and other image manipulation code in aforge. Together with alot of other cool stuff like machine learning and fuzzy logic etc.

Aluminum frame with the mount on top

Sun, 14 Aug 2011 14:10:27 +0200

Aluminum frame with the mount on top

I really want to learn how to solve this little monster.. one...

Fri, 12 Aug 2011 17:22:01 +0200

I really want to learn how to solve this little monster.. one side! Six to go… (Taken with picplz.)

Bored (Taken with picplz.)

Thu, 11 Aug 2011 21:33:30 +0200

Bored (Taken with picplz.)

“High” speed train (Taken with picplz.)

Thu, 11 Aug 2011 21:28:49 +0200

“High” speed train (Taken with picplz.)

School project: autonomous sentry gun

Tue, 09 Aug 2011 18:15:00 +0200

In about three weeks I am heading back to school. The first month of the next semester is focused on two of three subjects. The three subjects chosen by students is

Android development

Game development, using either xna or unity.

System administration.

Then we are supposed to pick a “end product” to finish during the month. We are then to select one or at most two of the subjects most relevant to our final product.

The final product could be a RSS reader for android. A xna game for windows phone 7 etc. A while ago we voted on which subjects we would like to attend to. And unfortunately embedded systems was not among the final three subjects.

I had been collecting parts for a year or so, to build a sentry gun, like the ones you see in a game, like team fortress or black ops.

And I told the teacher about it and he gave the green light on me building that as my final project.

I found that the most relevant subject is game development, so I chose that, and then I am going to assemble a overview of articles, tutorials and presentations on motion detection, AI, and a bunch of other stuff, I can use when programming the final product :)

Would you understand me if I said I can’t wait to go back to school? :D

Weird error in WPF

Tue, 09 Aug 2011 17:05:00 +0200

I have a class that imports some functions from a c++ dll file, using the C# keyword extern.

The class controls a webcam, contains an event that fires when a new frame is received.

In a C# console application it works great.
And in my WPF project, i created a small prototype that works aswell.

In the prototype the viewmodel is bound like this

Remember, the above way is the one that works.

Now i started implementing it in my main application, and best practice was to bind using the ViewModelLocator, and set the path like this

When i use the binding that’s using the ViewModelLocator, the camera class is retuning nothing but a black image, and it’s framerate lowers to the timeout value, so something is messing up somewhere inside it.

So the solution to my problem was to just use the “direct” binding, the one without the ViewModelLocator.

Now, i have no bloody idea why the other way of binding the viewmodel to the view is working, and i would love to hear a reason if anyone knows.

Now, this post is not just for my own benefit, if this or some of the following posts can help others, I will indeed be happy.
I just hope people will be able to find this post with the lack of keywords and proper description. English is not my main language, and even if it was, exactly what is this error called?

Don't ruin an apology with an excuse

Tue, 09 Aug 2011 17:04:57 +0200

After thinking about the meaning for a minute or so, it makes perfect sense.

An apology is something you offer someone you hurt in some way. You are telling someone that you are either sorry about or regret something,

“I’m sorry about being jealous and keying your car”

That’s a genuine apology, but try adding “but you were really mean”. Now, not so much.

There are apologies that this doesn’t apply to, like

“I’m so sorry about crashing your car, but that aquaplanning was really crazy”

But it does apply to every apology concerning something you did on purpose.

Random thoughts ftw :)

Errors and solutions

Tue, 09 Aug 2011 17:03:55 +0200

When I stumble upon an error I usually find a solution. And after that I always think about writing the solution down somewhere, so I won’t have to go through the same steps the next time I’m facing the same error. But I always end up thinking “yeah, I should do that”… then nothing…

Until this very moment!