in the wee hours

The Archives

  • in the wee hours
    Hitting the ball Uncategorized | (0)
    My "intercept" module is starting to produce results -- at least in this simple "literal" use of its features. Spent some time over the past couple days clearly separating out the "guidance" from the "intercept" part of the code. This is one foundation idea from the general area, and turns out to be a good one for getting things to work fast. Basically the s/w is divided into those parts that "simply" (yes, I'll put that in inverted commas) guide to bot to some location given a set of parameters, and those parts of the s/w that determine the point ...
  • in the wee hours
    When good VisionServers go bad… Uncategorized | (3)
    The other day I decided to have a look into using the OpenCV vision library in a couple of my projects. If you have never heard of it, OpenCV is a free, high performance vision library that was initially developed by Intel and has since been open sourced. The lib contains a lot of handy algorithms that allow a developer to fairly easily solve vision problems like object detection, feature tracking, image analysis and segmentation. The lib comes in various modules with one containing high level algs, another containing low level operations and another containing high level windowing functions. As I have ...
  • in the wee hours
    Gesture controlled robot Uncategorized | (3)
        Check this out. Tokyo University engineer Tsuyoshi Horo has developed a new body gesture robot control system. The robot itself in the videos is nothing too interesting (although I would love a robotic chair to drive me around my house) rather it is the fact that he has managed to fuse the image data from multiple cameras into a 3d voxel map and then used the gestures from that to control the bot. Hmm... now to implement a similar system in my house :) Check it out here: http://www.engadget.com/2008/02/19/gesture-controlled-robot-is-at-your-service/
  • in the wee hours
    Robot Tennis Uncategorized | (0)
    With some progress made on the Bioloid simulation front, I'm now trying to add a few more capabilities to the basic robot s/w. Being able to solve a few static scheduling/constraint problems is one thing, but RL is not static. The planner that's central to most of the robot operations I've been playing with so far (distinct from using robots that are not much more than tape recorders) has been purely "static". It has no in-built concept of time (although that can be added via constraints that explicity mention and maniuplate a "time" item in the state), nor associated ideas ...
  • in the wee hours
    VR Spider Bioloid almost working Uncategorized | (0)
    Some late nights over the weekend has seen some problems with the animation s/w worked out, and at least a lead on another problem related to the "s curves" highlighted. The basic Bioloid program is now running in the simulated space, albeit the physics is still a little funny-looking. E.g. the floor is just a little too elastic, and you can notice the robot bounce from side-to-side (a little) as it uses its 2 "side feet" in some of the free movement clips. The attached vid shows (although not explicitly) responses to sound and light stimulation. While in "free roam" ...
  • in the wee hours
    Caching = true Uncategorized | (0)
    Hi guys, I have turned on the caching system for the site so hopefully page loads will be a little faster. The Drupal 5 caching is pretty simple however so don't expect too much. Drupal 6 has recently been released with improved caching support so I will upgrade the site to that as soon as I can guarantee that everything won't explode. :)
  • in the wee hours
    King Spider animation turns up a problem Uncategorized | (0)
    It's been said before many times and better. But I'll say it again. Systems of even moderate numbers of simple rules are hard to understand. Since I don't spend my time writing "first order code" -- i.e. code that just does a given job -- many parts of my animation s/w are presently generated by various tools from simple specifications. Prolog is good for such tricks. The module handling the location of the robot in the V.R. is one such module. Knowing simple things like "if the robot keeps all its feet on the ground then it can't change its ...
  • in the wee hours
    Motion planners — one step forward, 2 steps back Uncategorized | (0)
    Another week gone and another minor version or 3 of some of the mid-level s/w re-written to eliminate gross and unsuspected bugs. :) After a while you become immune to wondering "how the Hell has this managed to fool me for so long it's doing something?". It's just the nature of s/w. :} Have exchanged brief pleasantries with a fellow working on a motion-designing s/w for Pleo. The last I heard on the dinobot front, there were still various "issues" relating to its release. But I have seen at least 1 civilian on YouTube that has one that appears to ...
  • in the wee hours
    Dean Kamen’s robotic “Luke” arm captured on video Uncategorized | (0)
    Wow I can't wait to see what this stuff is like in 50 years. I may even have to get my arms removed and a couple of 'em attached in their place. Can't let the young whippersnappers push me around as an old man! Now all we need is a better feedback and control mechanism and it will be perfect... Oh yeah and lasers. Don't forget the lasers! :)
  • in the wee hours
    Comparing simple zero-point gaits Uncategorized | (0)
    Apart from debugging 3d modelling and graphics rendering code, have made some progress in the problem I got into this whole business to sort out -- trying to make my quadbot run. To recall: in the model I built using the Bioloid kit, I'd *observed* it what I call "ninja running" across the carpet for short periods of time. (I.e. for 5-20 steps it would use a "galloping" leg movement pattern with straight legs and make very rapid progress). I've not been able to isolate HOW this is happening, apart from noting the "real world" s/w I'm using has both ...