'Techno-maniacal-freakishness'

Pipelining image processing with OpenCV in a GUI (GTK+)

noreply@blogger.com (Sudeep Pillai) — Wed, 11 Mar 2009 08:28:00 +0000

Lately, I've been kept recluse by the very thought of a machine vision application that will let me rapidly develop image processing techniques via a GUI interface. Obviously, now that I've been enlightened by the power of OpenCV, I'd want the GTK based GUI application to stack filters one on top of another so that an image processing pipeline can be quickly and easily established. This also solves the whole issue of writing code for individual pipelines especially when dealing with several images/feeds at once.

Roborealm does exactly what I've described except in a different image processing language, and is currently well documented/established. However, I did want the ability to write my own filters in OpenCV specifically so that I can further understand machine vision like I've always wanted to.

Virtual workspace for optimized performance

noreply@blogger.com (Sudeep Pillai) — Thu, 05 Mar 2009 05:41:00 +0000

Here's a wacky idea that's sort of parallel to the Bumptop project. What if we could interact with our computer the very same way we interact with our environment?

Visualize a virtual workspace if you may, a study room with a desk in front of you. You have your typical items/devices on your desk, an LCD monitor (possibly more than one), a bunch of documents that require your attention for that current day, a trash can to throw out those redundant memos and mails, a shelf on the side to organize your daily reads, a todo cabinet/folder/stack and so on. Now, what if your computer can organize all these items for you so that all you need to worry about is finishing the task at hand.

Some cool features may be:
- You can set the order in which you need certain documents/tasks to be done by physically moving those documents from your "computer directory" to a virtual area such as the todo stack that sits right next to your virtual desk.
- Keep track of what you've accomplished as you see the todo stack growing/diminishing in physical size.
- Have a music player on the other end of the room, all the music can be categorically organized. Visual interaction can be effortless and elegant.
- Design your own workspace to your comfort (possibly using Sketchup, simple and fun to use)
- Share these workspaces between people (shows you how well others organize material, and accomplish tasks)

I'm sure there are plenty more possibilities with such an interface, however making it visually appealing can be a challenge.

Useful desktop notifications on Ubuntu

noreply@blogger.com (Sudeep Pillai) — Thu, 05 Mar 2009 03:25:00 +0000

I've been a loyal ubuntu fan for over 2 years now. With every linux distribution comes the power of customization which I recently discovered was my main argument for loathing windows and embracing linux in the form of Ubuntu. Anyways, I've recently been tinkering with the desktop notifications provided via libnotify. The feature of an unintrusive notification system can be taken to great lengths by incorporating some intriguing ideas.

Some interesting tweaks/implementations I currently enjoy with this feature are:
- Listening in on tweets
- IM notifications

Some features that may be fun to implement:
- Messaging between local computers. Technically the message self-destructs in 5 seconds.
- Gmail notification/message previewing
- RSS feed reader (currently I use it in the form of tweets of blogs i.e. engadget, lifehacker, nytimes)

Mini-blogging initiating... this very moment

noreply@blogger.com (Sudeep Pillai) — Thu, 05 Mar 2009 02:12:00 +0000

With twittering doing most of the "long-story-short" talking, I thought I'd just make my blog posts a bit more easy to keep up with. Even listening in on tweets from a bunch of people can be difficult. I leave it to my GNOME-Do to notify me of all the tweets via libnotify. It's simple, uninstrusive and best of all, I can visit the links only if the title of the message is interesting.

Let's face it, no one really has any time these days to read random posts and expect it to be a worthwhile read. Keeping that in mind, all my subsequent posts will hopefully be concise.

P.S. Take a look at who I'm following and those that are following me on my twitter page.

Poor man's GPS (redux) on PyS60

noreply@blogger.com (Sudeep Pillai) — Fri, 16 Jan 2009 20:22:00 +0000

Lately I've been enthusiastic about the idea of scripting on my Nokia-N70 using PyS60. Writing programs on these phones has never been easier once Nokia released python bindings for their Symbian C++APIs. Anyways, I've been thinking of techniques to salvage the most of technology in to the palms of my hand, much like every other mobile software company, and I came up with this.

Among the myriads of technologies out there, I wanted to mimic 'MyLocation' - a recent update by Google Maps for smartphones that approximates your current location based on the cell tower you are currently connected to.

MyLocation on Google Maps

Here's how it works:
1. Your phone is connected to a particular cell tower and has certain cell tower parameters that it can access (here's where the PyS60 API would come in - look for location.gsm_location() in the code)

2. Once these parameters are collected (mcc - Mobile Country Code, mnc - Mobile Network Code, lac - Location Area Code, cellid - Cell Tower Id - read GSM Localization), Google's 'secret' API retrieves the longitude and latitude of that particular cell tower based on the tower parameters.
Note: The longitude/latitude retrieval is accomplished by packing the data in a specific manner (review code for this).

While this is amazingly helpful in a strong network neighborhood such as Manhattan, it usually lacks the accuracy (errors of upto half a mile) in sparsely networked areas. Google claims that they somehow triangulate the positions of the cell towers and provide you with the best estimate of your current location. However, the abilty to retrieve cellid's from different cell towers is purely handset-specific and most handsets only report the cellid of the cell tower that provides the best signal. This seems more of a software ended issue where the phone decides to automatically choose the cell tower that provides the most signal power thereby keeping the phone 'reception' high.

Now here comes the 'hack'; what if we could ping all the cell towers within the neighborhood for a very small period of time until each of their GPS coordinates are retreived. Triangulating these coordinates can further result in a more accurate position estimation, specifically useful in navigating streets etc. This method does however require the phone to drop the connection with its current 'high reception' cell tower in order to ping the rest of the cell towers.

Another 'tweak': It'd be interesting to run this while you are cruising down an interstate especially since there would be only certain towers that provide strong signal strength to those areas. This y we could better approximate our current position and also be certain that we are on the interstate as opposed to 'somewhere in this vicinity'.

Here's the code I used (originally written by Neil Young - slightly modified by me) to retrieve the GPS cooridinates via google maps on my Nokia N-70.
Download

PyS60 project website can be found here. [contains manuals, installer files etc.]

Stellated Rhombic Dodecahedrons

noreply@blogger.com (Sudeep Pillai) — Thu, 01 Jan 2009 08:26:00 +0000

This brings together the best of both my worlds - math and geekiness. Enjoy!

If you liked the above, this next video shall also amuse you. It's called a Gömböc. It's a mathematical class of objects with one stable and one unstable point of equilibrium. This means that regardless of how you place it on the ground, it continues to self-right until it reaches a particular orientation.

Happy new year everyone!

Billboard Top 25 of 2008 in '4 minutes'

noreply@blogger.com (Sudeep Pillai) — Wed, 31 Dec 2008 20:29:00 +0000

DJ Earworm brings an interesting mashup of all the popular songs of 2008 condensed in a span of 4 minutes. The best part is that it's legal to download it.
Download: "United State of Pop 2008"

The mash-up currently mixes the following songs in a relatively pleasing fashion:

Flo Rida Featuring T-Pain - Low
Leona Lewis - Bleeding Love
Alicia Keys - No One
Lil Wayne Featuring Static Major - Lollipop
Timbaland Featuring OneRepublic - Apologize
Jordin Sparks Duet With Chris Brown - No Air
Sara Bareilles - Love Song
Usher Featuring Young Jeezy - Love in This Club
Chris Brown - With You
Chris Brown - Forever
Ray J & Yung Berg - Sexy Can I
Rihanna - Take a Bow
Coldplay - Viva La Vida
Katy Perry - I Kissed a Girl
T.I. - Whatever You Like
Rihanna - Disturbia
Rihanna - Don’t Stop the Music
Natasha Bedingfield - Pocketful of Sunshine
Chris Brown Featuring T-Pain - Kiss Kiss
Ne-Yo - Closer
Colbie Caillat - Bubbly
Mariah Carey - Touch My Body
Madonna Featuring Justin Timberlake - 4 Minutes
Pink - So What
Finger Eleven - Paralyzer

If you found that interesting, then go ahead and check out the 2007 mash-up.
Download: "United State of Pop 2007"
There's plenty of mash-ups on this website that is worth listening to.

BOIDS.. Nature always finds a way

noreply@blogger.com (Sudeep Pillai) — Tue, 30 Dec 2008 08:49:00 +0000

Here's a simulation I did (for my Intro to Adaptive Systems course) on bird flocking while following some simple but decentralized rules. Apart from following the simple rules of Separation, Cohesion and Alignment, I've also added some goal tending influences so that the organisms tend towards the circular object. All the code and simulation was run on Matlab using some slightly modified code from Conrad's pseudocode for boids. I've also linked the code at the end.

Boids simulation on Matlab from Stud Sudzzz on Vimeo.

EE594 Intro to Adaptive Systems - Presentation on Boids
EE 594 Intro to Adaptive systems - Final paper
CODE: Flocking code in Matlab

Project Timeline

noreply@blogger.com (Sudeep Pillai) — Tue, 17 Jul 2007 04:54:00 +0000

SPARCCA 's the name. Here's how it grew.
SP - Sudeep Pillai, A - Autonomous, RC - Remote Controlled, CA - Car (shortened)

4/10 Initial concept about AGV's and their implementation on a smaller scale

4/24 Pitch idea about an autonomous RC car as a side project

4/24 Professor Ryan Eustice requests to focus on this particular project if genuinely interested

4/24 Request for some sort of funding from the College of Engineering (Really doubtful though)

4/25 All funding opportunities and positions are taken. No funding would imply a smaller project with a constrained budget.

4/26 Another e-mail from Prof. Ryan Eustice. A Summer Undergraduate Research Award (SURA) position opens up. I got it! I'm absolutely elated. I start delving into the world of autonomous vehicles. Its just a perfect break I wanted to prove myself to the world.

5/9 (Blog) My first blog entry

5/12 (Blog) My masterplan - where I unveiled my project details and my endeavors

6/4 (Blog) Purchase of components - Wireless camera, USB TV tuner box, RC truck and 12V battery for portable power.

6/4 (Blog) Initial Design Document (doc) - A compilation of my project tech specs, goals and objectives.

6/7 (Blog) Programming Languages Dilemma

6/12 (Blog) MEX Functions and MATLAB

6/17 (Blog) Machine Vision Updates

6/19 (Blog) MATLAB & C++ interfaced with RR

6/28 (Blog) RC controller with PC link

7/9 (Blog) Lane detection and Terrain finding

This timeline still requires filling in the gaps. It should be done in a couple of days. Please be patient. Thanks.

Lane departure systems terrain finding

noreply@blogger.com (Sudeep Pillai) — Tue, 10 Jul 2007 02:05:00 +0000

After a very lethargic week, I've propped myself right back up into 5th gear. It took me some time to convert the strings from RR (mainly from the hough line data) into sensible data so that they could be used for further processing in MATLAB. The data goes into MATLAB as hough_lines and are used by some specific systems that I'm currently working on such as a lane departure system(its quite interesting how the same system* is implemented in the Lexus LS460** and Citroen***). Furthermore, they will also be used to compute different elements to facilitate splitting the video feed into various ROI's (Regions of Interest) which would allow for easier computation by reducing the total number of pixels for image processing.

Here's the plan.

The lane departure system and terrain finding will be split into several different systems. First, we deal with the visual feed and how we can interpret the data into sensible data for lane and terrain detection.

[+] With the visual feed, I plan to split it into 3 different ROIs.

[-] The first one (ROI1) will occupy the lower area of the video feed spanning from the bottom till y = 60 pixels. This area calls for impulse reactions from the car and hence I chose to implement a center of mass (COG - Center Of Gravity module in RR) calculation so I can figure out where exactly center of the road lies.

[-] The second one (ROI2) occupies a dynamic area which changes based on certain hough characteristics. The area is dynamic since 'ycrop' which defines the lower bound of this region is a variable which changes with time. This variable is retrieved from the hough transforms created by edge detection of horizontal lines or those that are angled between -15 and 15 degrees. The higher point of the hough line that satisfies this condition is taken as ycrop. Here's an example of how the hough transforms and the region will look like.

[-] The third region will occupy most of the area spanning from the bottom till ycrop, the dynamic cropping variable previously mentioned. Here most of the hough transforms will take place and this will allow to predict the direction of the upcoming terrain and align the car accordingly. One obstacle that I encounter is that the hough transforms give me skewed results with the change in light intensity. This can be corrected by extending the hough transforms formed by the lanes to intersect within a bounding box. Through this, the significant('base') hough lines can be determined. Any other hough line will be disregarded. Here's yet another picture to illustrate it. NOTE: This is applicable only to considerably straight roads.

The combination of these 3 regions and their computations that would take place will provide a reasonably robust system for lane detection and car alignment or maneuvering if you will.

So far its been nothing but spending 8-10 hours in front of the computer trying to figure out how to relay data between MATLAB, Roborealm and the C++ code. I should have a working model by the end of next week with another week of coding the AI where the RC car would run solely on vision. With just hough transforms and center of gravity modules the car should be able to encounter curved terrain as well as road edges so that it keeps itself within the terrain. I'm gonna meet up with my professor, Ryan Eustice (I have to say, great guy. Really helpful.) later today and will possibly establish that I would want to start up a project of my own. The plan is to start small so that most of my ideas are implemented before it gets mixed up in a larger crowd of people. In terms of the demographics, I hope to have people from different fields of expertise such as computer engineering, electrical engineering and mechanical engineering. Maybe its a bit too early to be dreaming about my a student group led by me but its a worth a try.

Adios

* The system mentioned is Lane Departure Warning System (LDWS) , LDWS@cnet-review
** The Lexus LS 460 LDWS a.k.a Lane Keeping Assist
*** Citroen's LDWS

Google buys GrandCentral - take that iPhone visual voicemail

noreply@blogger.com (Sudeep Pillai) — Tue, 03 Jul 2007 22:00:00 +0000

A recent and well thought out move by Google to buy Grandcentral, a VoIP based company, possibly stirred the geeks at Apple and their fan club as they unveil certain features to allow a central link to all your phones.

Check it out. Apparently they are offering free phone numbers. Just a rumor, yet.

http://www.grandcentral.com/

iPhone - A great marketing campaign, an ordinary phone

noreply@blogger.com (Sudeep Pillai) — Fri, 29 Jun 2007 03:53:00 +0000

The iPhone is out tomorrow at 6pm. After months of much anticipated wait, it finally arrives tomorrow. But painfully all that wait has gone in vain for it doesn't prove to be a GREAT phone for the future. Its just something that you could have for a year or two but eventually give in to those 3G phones out there. And for a steep price of $500 AND a two year contract from AT&T, (wait..what? $500..yes that too but, what company.. AT&T..what's that again? T-mobile you say.. no, AT&T..Cingular??..no, AT&T..huh??.. AT&T - The Leading Cellular Service -- In Customer Complaints)

What Apple did was epic, there's no questioning about their marketing campaign. They thought, 'hmm.. How could we possibly conquer the cellular phone market.. I know.. use our most powerful weapon against them.. the iPod.. add a couple of gizmos and our OS X' to create a great market competitor.

'Need For Speed' for the real world

noreply@blogger.com (Sudeep Pillai) — Fri, 29 Jun 2007 03:07:00 +0000

Yesterday was a momentous day. I was too excited to even post a blog coz I was playing around with my real life hummer H1 using my computer. Yes, I've finally setup a PC controller for my RC car and have been messing around with the signal pulses that it sends. Granted that the car didn't respond too well initially, it just needed a few quick adjustments in the signal transmitting code to produce a smooth acceleration in the car. Furthermore, I needed to figure out a more efficient turning mechanism because like many I/O boards the response usually requires a small delay in the pulses sent otherwise the board shorts and asks for re-initialization.

Having progressed this far I decided to cut myself a bit of slack, just for a day though, and hooked up my wireless camera with the car and setup a life-like car simulation where I tried to maneuver through chairs, tables and boxes that were carelessly strewn about in my apartment. The feeling was quite exhilarating and the sight that RoboRealm provided was remarkable; except for seeing my roommate walking out of his room wearing only boxers heading for the shower. Anyways, this week has been very productive for me having accomplished several little tasks that were pending for quite some time. I've almost setup a GUI interface that shows me the different video outputs that RoboRealm processes and have most of the commands running from a single program as opposed to manually running each program separately (RR, MATLAB code and GUI interface, C++ code).

Now its finally time to conjure a brain for the car that can make it a 'treat' for those geek news sites, tech blogs and the 'geeks' obviously. Keep commenting and expect more frequent updates.

Finally.. MATLAB & C++ - Unleashing their powers

noreply@blogger.com (Sudeep Pillai) — Wed, 20 Jun 2007 03:09:00 +0000

Its been a productive day for me after having figured out how to control MATLAB with the help of C++ functions. With the use of MATLAB's inbuilt Engine library you can control, edit and run programs and variables through a C++ program. This way I finally have an intermediary through which I can communicate with RoboRealm back and forth allowing me to effectively bring together the visual perception system a.k.a "The Eye" and the Artificial Intelligence (AI) system a.k.a "The Brain". The C++ program acts as an "optic nerve" catering to sending and receiving data to and from MATLAB and RoboRealm. As for now, I've been trying to receive image files from RoboRealm through its API server and manipulate them before I continue sending it to a MATLAB array at a sampling rate of around 3Hz. This low sampling rate is more than enough for inspecting terrain changes and detection. Once all the data transmission systems are complete I should be ready to take on the huge task of programming the AI system.

Check this webinar out!
MATLAB for C/C++ Programmers
Its really helpful if you wanna use MATLAB's engine library with C++.

Machine vision updates

noreply@blogger.com (Sudeep Pillai) — Mon, 18 Jun 2007 02:36:00 +0000

After having tinkered with MATLAB for quite a bit of time since last week, I've gotten to a point where most of the programming involves optimizing the code for trajectory following. Initially I came up with a bare bone architecture for path following and obstacle avoidance. Currently I've been working with making a more robust system for tracking the edges of the traversable path and thereby determining a corresponding trajectory for the vehicle.

Here is an overview of my progress in the past week.

[+] Used RoboRealm (RR) to setup a robust vision scheme for optimally tracking the edges of a traversable region and finding each of their corresponding slopes for further application development.

- This first image shows a snap of a video feed that is filtered with RR using a couple of its inbuilt filters. (Picture - Right)

- A color threshold is added to the above image to obtain a video feed that only shows the path that has to be traversed. (Picture - Left)

- A Convex hull algorithm is added to the above setup to obtain continuous curves and edges. This allows an easier implementation of edge finding and allows faster image processing in MATLAB which will be explained later in this post.
(Picture - uhmm.. You should have figured it out by now)

- An outline of the convex hull algorithm.
(Picture - Oh, cmon. Are you kidding me?)

- Hough transform extracted from the above image. This image is grabbed by MATLAB with the use of MEX functions from the Roborealm API server (still need to figure out how to do this. Please let me know if any of you have any clue).Currently, its just an image read from the local disk.

[+] A MATLAB algorithm that efficiently finds the midpoint of the two edges and thereby creating a smooth trajectory for the vehicle to follow. The m-file uses a hough_eqn function that is also a user defined function which finds the equation for the hough edges using the data that is provided by Roborealm such as x,y coordinates of the endpoints of each line.
(The algorithm is still in beta testing mode and will be uploaded once completed.)

The midpoint trajectory algorithm is quite efficient primarily because the all the pixels in the image are not accessed. This is because, with a given set of coordinates for the continuous lines (they have to be continuous otherwise the algorithm will be inefficient) the algorithm determines the possible pixels that contain the lines and thus eliminates unwanted access of pixels that don't contain any information about the edges. Although accessing each and every pixel in the image is eliminated, an additional task of finding slopes of the hough lines is added to predict the direction and location of the lines. However, this additional information (slope of the lines and location) will help the vehicle to determine how straight its trajectory is with respect to the path.

[+] Another progress I've made over the week involves coming up with a crude method for obstacle detection purely based on vision. Further info shall be provided in the next post. Please leave me comments if you would like a more detailed explanation of any of my algorithms or any particular aspect for that matter.

Another update..Well not much of an update..but i call it one

noreply@blogger.com (Sudeep Pillai) — Tue, 12 Jun 2007 06:00:00 +0000

After a long and anticipated wait, the RC vehicle finally arrived today. It took them more than 10 hours to deliver the package from the time it left the service facility.. pssh..and they left the package at the doorstep without even notifying me. Anyways, I walked down eagerly to see one of the biggest packages I've ever received. The whole thing was about waist high (1.5ft x 1.5ft x 3ft). Anyways, after lugging it up into my room I unraveled a monster of a vehicle which could technically give rides to toddlers . It was definitely big; bigger than I expected actually. However, come to think of it, I could use its sheer size to add more components and sensors as I wish and never worry about whether they would fit on it or not. As for my work, I've still continued to show considerable amount of frustration because of all the different aspects of the project and piecing them together is quite a task. After having posted a question on the Roborealm Forum asking for help on MEX functions (MATLAB based functions that allow c++ programs to access API's, those of Roborealm), it's been a painful wait. This way I can start manipulating 2D images that are captured by the camera and convert them into a scaled 2D map (as mentioned in an earlier post) of the surrounding environment.

Couple of interesting links in the past week that caught my eye
MIT wirelessly powers a light bulb
Apple Safari On Windows Broken on First Day

I also wanted to take this opportunity to thank all those who read these posts and would love to hear your comments and suggestions.
Thanks, once again.

Programming languages - Choosing the right one

noreply@blogger.com (Sudeep Pillai) — Fri, 08 Jun 2007 03:24:00 +0000

After a painful day of exploring the different programming languages that I could use(VB, VC# and MATLAB) and their available support for video for windows (VFW), I have come to a difficult stage of my project where I 'decide'; decide between two very robust yet different programming environments, VC# and MATLAB. Having to decide between the two languages brought me to one of the very fundamental questions a researcher would ask him/herself ? Do I write in a language that gives me an easy and efficient method to test out my algorithms, be it for navigation of AI, or do I write in a language that gives me the power to expand my software based machine learning environment which will allow to create an RC car with a powerful brain to traverse unconquered territory( just kidding).

I'm losing it

noreply@blogger.com (Sudeep Pillai) — Thu, 07 Jun 2007 03:23:00 +0000

So, I've been trying to figure out a way to pipe the processed video feed from RoboRealm into a VB program which will be used to create a 2D cost map for navigation of the vehicle. And sadly it's been taking me a painfully long time to interface the two (RoboRealm and the VB code). I've spent most of my time for the past few days learning new scripts and VB language and have failed to accomplish my task (well, as of yet). I spend around 10-11 hours on the computer nowadays browsing forums and message boards trying to figure out an efficient real time image processing code in VB.

Apart from all the frustrations I've been excited with all the packages (components mentioned earlier) that are being delivered.

Highlighter range finding

noreply@blogger.com (Sudeep Pillai) — Tue, 05 Jun 2007 18:20:00 +0000

After fiddling around with Roborealm, an open source machine vision software, I've made a simple range finder only with the help of vision. Granted that the visual feed is still and that it requires a reference frame that is also relatively stationary in the x-axis, I've setup an easy method to find the range of an object. Here, I've used a highlighter because it stands out from the surroundings and thus easy to track. As one can see from the video, I didn't have a scale of measurement and therefore used a regular A4 sized sheet (which I stole from the computing lab) to measure distances. The range_ratio is the unit of measurement (single A4 sheet) and a range_ratio of 1 corresponds to 297mm in length.

The range finding also involved calibrating the pixels in the visual feed into real world units of measurement and this required writing out a calibration graph.

Here are some pictures of the notes I took to supplement the above.

I ended up with a reasonably good approximation of the range with this formula:
Distance = 2 *Actual distance/# of Pixels

# of Pixels was computed as xdiff in the video

Please let me know if there are any questions.

Project Update - Purchase of components

noreply@blogger.com (Sudeep Pillai) — Tue, 05 Jun 2007 00:12:00 +0000

After having hogged down mum's mouth watering food and having spent some quality time with the family, I find it very difficult to get back to the same old pace at work. Nonetheless, I've recently purchased some components (which sadly hurts my own wallet coz the department still hasn't mailed in the checks) allowing a smoother and faster progress of the project. Anyways, here's a list of items I purchased.

1. Wireless Infrared Camera / Color Camera
- Operates at 2.4 GHz
- Illumination: 1 Lux ( f:2.0) Day Light and 0 Lux at Night Light with infrared LED On (30 LED elements)

2. 12V 4.5 Amp Li-Ion Rechargeable Battery Pack with Charger
- For the wireless camera to allow remote operation

3. PROLINK TV-USB-2.0 TV BOX
- To allow connectivity between the laptop and the Wireless Camera

4. 1:6 Scale Remote Control Hummer Monster Truck
- Chose this particular model due to its high ground clearance

To have a better understanding of all the possible components that might go into this project, please read the Initial Design Document. Although it isn't as detailed as a technical document, its gives the gist of the entire project and my future endeavors.

Steve Jobs and Bill Gates: Historic discussion live from D 2007 - Engadget

noreply@blogger.com (Sudeep Pillai) — Sun, 03 Jun 2007 16:06:00 +0000

I apologize for the dormant blog that has been up for the past couple of weeks and would like to update it more often, hopefully everyday, with my project events and details as it progresses.

One interesting article (or video) is an interview between two of the most influential people in the world of software and need no further introduction; Bill Gates and Steve Jobs. Check this one out, these don't show up every so often on tech sites. I couldn't help but steal this momentous quote from Engadget, "These two guys are one in a million, and it's totally clear they've never respected anyone else quite like they respect each other."

Steve Jobs and Bill Gates: Historic discussion live from D 2007 - Engadget

Weekly Update

noreply@blogger.com (Sudeep Pillai) — Sun, 13 May 2007 20:41:00 +0000

This is a bunch of articles that I found interesting during my regular read of tech sites.

A $10 Laptop.. Huh?
A brave and passionate move by Nicholas Negroponte (OLPC - One Laptop Per Child) a couple of years ago to build a notebook for $100 has recently been slashed by a group of researchers from the Indian Institute of Science and Vellore Institute of Technology claiming that they could make a laptop for a fraction of the cost.

Thought: Indian masterminds to the REAL rescue!

AquaSkipper - Human powered water vehicle
Pretty impressive for the amount of effort you put into it

Current Read: Probabilistic Robotics by Sebastian Thrun, Wolfram Burgrad and Dieter Fox
Current Song: Hallelujah by Rufus Wainwright
Current Food: Tuna Melt

My 'masterplan'

noreply@blogger.com (Sudeep Pillai) — Sun, 13 May 2007 04:29:00 +0000

As I mentioned in my previous post, I shall now disclose or may I say 'unveil' my master plan for the summer. Recently, I've been fascinated with the idea of autonomous driving and computer assisted driving even though it has long been a fascination for many. One particular event that caught my eye was the DARPA grand challenge that took place in October 2005. It was a initiative undertaken by the US military to build autonomous ground vehicles primarily to simulate unmanned supply missions (well at least that's what they claim).

Having read a lot about this I looked into my university that affiliated with something that was even remotely close to this project. I found a professor who was working with one of the local teams that was entering the upcoming DARPA urban challenge (which will take place this November in the city). He worked with the integration of LIDAR (Laser Detection and Ranging) in to the vehicle for obstacle and terrain detection. As the project progressed I mentioned to my professor how I planned to build an autonomous car based off of the knowledge I received from the research project. And now, a couple of weeks later, I've been alloted a sum of $3000 to invest in my very own research project with my professor as my mentor.

My main goal is to build an autonomous RC car (if that makes any sense at all; basically an autonomous vehicle on a smaller scale of say an RC car), that is capable of recognizing terrain patterns and obstacles so that it can maneuver itself autonomously. This project goal can be recognized only with the implementation of clever vision technology and sensors so that they provide a reasonable estimate of the surrounding environment. Currently, I've been researching on the most suitable method of terrain recognition and found RoboRealm, an open source machine vision software, to be extremely useful. Even though I haven't tweaked the program to utilize its sheer image processing power, I've successfully configured it to track the floor using pattern recognition so that the vehicle can distinguish between drivable and non-drivable regions. I've also considered the various technologies that were actually implemented in the previous grand challenge. One particular technology that I wish to use is the 'adaptive vision' that was incorporated into Stanley, the Stanford Racing Team's car that competed in the challenge and ended up winning the race. It uses clever terrain recognition using a visual feed, say a video camera and tracked the terrain that was closer to the horizon and matched it with the terrain that was immediately in front of the vehicle at that particular instance. If the terrain matched, then Stanley was configured to accelerate. Otherwise, it would decelerate.

Having said this, I plan to incorporate several other techniques into my vehicle to make it as robust as possible. I'll make sure to update my blog as I progress with my project and would greatly appreciate comments and feedback from all the readers (Thanks!).

My very first blog

noreply@blogger.com (Sudeep Pillai) — Wed, 09 May 2007 19:29:00 +0000

This being my very first blog, I think I should keep it short and simple.

I'm a junior at the University of Michigan - Ann Arbor pursuing my degree in mechanical engineering despite my strong inclination towards Robotics, Artificial Intelligence and Mathematics (don't ask why). I've decided to set aside some time during the summer to make my own blog so as to publish my thoughts and ideas globally as opposed to having it established only among my network of friends and faculty.

In my next blog I shall disclose my research interests and the research project that I'm working on over this summer.