Short update, the project is still alive, though stalling a bit, because I try to interface hbot with the linux bluetooth stack (bluez), and it is far from being easy. Bluez needs dbus, dbus has very low-level interface so you need a binding like the ones available in GLib and Qt, but I don't want to depend on these (potentially huge) libraries. So I try to use dbus-c++, an almost-working C++ binding library (the parts that work are really nice by the way). So expect better news soon!

(This message got lost in the limbos of the interwebs for some strange reason, but here it is)
HBot v0.7.6 is out.

A lot of partially implemented stuff here: a new hardware interface started making its way to HBot. OSIF is an initiative for creating a open-sourced hardware, firmware and software for a servo-motor. The hardware controller used is a glorified USB-I²C converter, since the open-servos talk in I²C. For now, libOSIF is integrated to hbot source tree, and a dummy controller has been created around it. Some GPS support has been started too. Next version should see integration of bluetooth to hbot, so that bluetooth GPS connection burden will be off the system and handled by hbot. Lua bindings have been reworked (for a better integration). Finally, an Artificial Neural Network (ANN) library has been integrated, and can be used right now. It has the usual limits of ANNs, so don't expect miracles out of it; I could barely use it have simulated hbot learn an obstacle avoidance algorithm.

HBot v0.7.5 is out.

Some cosmetic stuff in there: “rare” 3rd party libraries are included in source tree for sake of simplicity, “SensorControl” changed its name to a more precise “AreaSensor”, introducing SensorArray and BumperArray, for Roomba first, we will see how player fits in that. The biggest move here is an effort to push down to the “senses” directory everything that is hardware-specific. This means that luabot, botmgt, as well all the algorithm directories will be hardware agnostic: if you add a new platform in senses, all the rest works out of the box. As a conclusion I (finally!) wrote down the luabot documentation, that you can find it here. Do not hesitate, subscribe to the hbot mailing list to get more details and give feedback!

As hbot grows, the needs to have more precise documentation, bug reporting and some discussion on new features arise. One of them is ready: an hbot-dev mailing list has been set up (thanks to Benoit), you can subscribe here. A second topic, which is the documentation, is on its way, using doxygen. You can find some references there already, it will need some polish and more effort to be complete (only luabot configuration phase is documented for the time being). Finally a bug tracker should be ready soon. Enjoy!

HBot v0.7.4 is out.

As previously said, there has not been much progress these last 2 months, so this release is merely to start new things on a fresh ground. The main point is the integration of OpenSteer, a library specialized in generating some movement behaviors for moving elements. It has been thought for virtuality, e.g. for creating a crowd that walks on the street, but with some adaptation I think it is possible to use it for some interesting behaviors in a robot, for example follow somebody, avoid obstacles, etc. The approach taken here enables several behaviors to be integrated into one steering. That is another thing brought into HBot in this version: the possibility to have several “move” events being integrated at the low-level movement manager. There is still a long way to go before the actual code acts like this short description (let's first make OpenSteer and HBot physical engine talk to each other), but I think it should be very exciting (if it works).

Happy new year, fellow roboticists! The project is on hold for 2 months now, because I changed job and country, and I am now living near Switzerland, Geneva area. When I get an internet access back I will be able to keep the project alive.

HBot v0.7.3 is out.

This new version brings a better TA* pathfinding, with funnelling optimization. The PID algorithm received much love, with a more generic approach, a better behavior regarding the different tasks asked to it, and some other niceties (e.g. bumper stops PID). There is also a functional OccupancyMapGenerator, that fills a map on robot move, with the trace of the robot. This will enable “blind” robots to create a map without sensors. Finally, some improvements in project code layout and genericity…

Progresses are slow, because the code at stake becomes quite complex and the problems to tackle are rather hard. Anyway PID moves are better, roomba odometry get better (yet not perfect), and a new map generator appeared: occupancy map generator, for marking as free space the places where the robot went. There is also some work in progress on making a beagleboard work, and preliminary investigations on the OpenSteer library (for autonomous move decisions) and OpenCV library (for computer vision).

HBot v0.7.2 is out.

Not much features, for this version prepared for a long time, but big features:

• native support for roomba 5xx: player was not satisfactory for real-time reactions (such as bumper reflexes), so I developped my own roomba driver, and it gives rather good results.
• pathfinding: an A* algorithm is used on the graph generated from vectorial map; results are OK, but the path could be simplified (with funnelling for example)

With HBot v0.7.2 you can use HBot-Eyes (aug09 version), which lets you control a roomba with a gamepad (start/stop the vacuum cleaner with pad buttons); a reflex handler written in lua makes the roomba go back when an obstacle is hit. It also makes several irritated beeps, and cuts gamepad for some seconds :)

It has been a long time from the last post. What kept me from progressing is some problem connecting the Roomba to my EZX phone, but this is solved now! I have been able to control the Roomba, pilot it and turn vacuum cleaners on/off with a gamepad plugged on the pyhbot machine. The control of the roomba is realized with HBot and player, but reactivity is not satisfactory. So I am currently writing my own roomba controller, to get a fine-grain feedback from the Roomba. As soon as I can, I post a video demonstrating HBot on Roomba. Stay tuned.

Find out what's coming in hbot!
This plan will change regularly to show project advancement.
You can also track wiki changes through RSS feed.

HBot v0.7.1 is out.

Introducing bumper management (very useful for the roomba, because it is almost the only sensors available) and more luabot interactions. Now you can write a signal handler in lua. More to come soon, for hbot 1st anniversary!

I received my roomba, and I have to say I am rather satisfied with it! I am already thinking about the ton of (not so) stupid things that can be realized with this object. On top of my priority list is to simply adapt basic hbot (as of v0.7.0) and command the roomba with a gamepad (with buttons for starting the vacuum cleaner). On bottom of the list (but high on the “fun” list), is to adapt a plate on it, and have it bring drinks and pringles when I invite people ;)

HBot v0.7.0 is out!

Work on SLAM has started, but needs a lot of tuning. I also focused on a new UI application: pyhbothbot-eyes. Distant pad control and display of map updates were needed to go further on the path to SLAM.
I am also waiting eagerly for a roomba I ordered a few days ago, let's hope it will fulfill both objectives of cleaning my room AND being hackable!

Oh, and… hbot is now beta! \o/

HBot v0.6.5 is out!

With this version comes a scripting API based on lua. Why lua? Because it is simple enough for everyone to try hbot without programming knowledge, it is easily embeddable, and it has a small footprint. Have a look at the luabot documentation. There is also a better integration to player/stage, now usable for testing without real hardware.
Finally, all dependencies are now optional; for example not having CGAL will disable vectorial maps, but the rest will be usable.
Enjoy!

I am starting this news topic to provide more information to anyone finding this page. Things are greatly improving lately, the integration of the player/stage project enable developping without actual hardware; the integration of lua provides an easy user interface; there will soon be an GUI application for better control of the mapping activity of the robot. Lots of software topics are to be implemented in the next version (v0.6.5), but there will be a focus on SLAM algorithms in a near future.