(inverse)-kinematics

Post all your hardware and software upgrades/projects here.

Re: (inverse)-kinematics

Postby roboalchemist » Tue Jun 18, 2013 10:39 pm

Michael,

Would you mind sharing the changes with me? I'd love to have them folded into mainline PoMoCo ;)

I just got samples of metal gears that do 180 degrees reasonably without jamming, and at a good price. It looks like we'll be offering a metal gear Hexy within a month or so (probably $300 vs $250 for plastic), and a discount bulk upgrade pack for people who've already purchased a Hexy.
---
Joseph Schlesinger - ArcBotics Co-Founder/CTO
roboalchemist
Site Admin
 
Posts: 671
Joined: Mon Jul 16, 2012 3:02 am

Re: (inverse)-kinematics

Postby rpcook » Wed Jun 19, 2013 6:55 pm

Michal,

Great example on the IK code to demonstrate the different sorts of walking gaits that are available with a hexapod. I'm very impressed. How long has this taken you, out of interest?

Thanks,
Rob
Rob
A record of my exploits and musings with Hexy and Sparki: http://robcook.eu, content available under a Creative Commons license.
rpcook
 
Posts: 121
Joined: Tue Sep 04, 2012 6:42 pm
Location: Luton, UK

Re: (inverse)-kinematics

Postby michal » Wed Jun 19, 2013 8:50 pm

@roboalchemist - will do, soon, probably this weekend.

@rpcook - thanks! The basic IK code took me about 2 days + about 14 days of polishing it (API, GUI, Joystick control etc).
michal
 
Posts: 36
Joined: Thu Jun 06, 2013 7:01 pm
Location: Delft, NL

Re: (inverse)-kinematics

Postby rpcook » Sat Jun 22, 2013 8:33 pm

16 days to get to this level of polish!? I really hope you are a student with little else to do with your time, because if you fit in this much awesomeness around a full-time job I'm going to have to come up with a new word to describe this level of productivity!
Rob
A record of my exploits and musings with Hexy and Sparki: http://robcook.eu, content available under a Creative Commons license.
rpcook
 
Posts: 121
Joined: Tue Sep 04, 2012 6:42 pm
Location: Luton, UK

Re: (inverse)-kinematics

Postby michal » Sun Jun 23, 2013 11:43 am

Well, I made the virtual (kinematic) robot and multi-point offset calibration before I started IK project, so technically I'm busy with Hexy for a bit more than month. I'm not a student for some time :-), but I use my brain a lot at work which helps me staying sharp.
michal
 
Posts: 36
Joined: Thu Jun 06, 2013 7:01 pm
Location: Delft, NL

Re: (inverse)-kinematics

Postby klims » Sun Aug 04, 2013 11:44 pm

Great explanation of the inverse kinematics Michael. Very helpful.

I have been trying to implement inverse kinematics on the Servotor32 for Hexy for a couple of weeks now.
Main trouble I think I am having now is nonlinearity of servos and math functions taking too long to compute on the microcontroller.

Has anyone else got IK working on Hexy? Or is anyone else working on IK for the Servotor32?
klims
 
Posts: 31
Joined: Thu Aug 01, 2013 1:31 am

Re: (inverse)-kinematics

Postby roboalchemist » Mon Aug 05, 2013 9:05 am

When you say working on IK, is there a specific problem to solve?
---
Joseph Schlesinger - ArcBotics Co-Founder/CTO
roboalchemist
Site Admin
 
Posts: 671
Joined: Mon Jul 16, 2012 3:02 am

Re: (inverse)-kinematics

Postby michal » Tue Aug 06, 2013 1:48 pm

klims,

servo nonlinearity is definitely an issue for accurate IK, but even one constant offset value per servo should be enough initially.

I agree that the math functions can be pretty demanding, but did you try to optimize the IK code? For example if you are trying to implement my IK code, than I would initially skip computation of body orientation - just keep it horizontal. Then you can store body points as a precomputed array. This already saves you about half of the computation time.
michal
 
Posts: 36
Joined: Thu Jun 06, 2013 7:01 pm
Location: Delft, NL

Re: (inverse)-kinematics

Postby klims » Tue Aug 06, 2013 10:53 pm

I'm working on something like what Michael has put together for controlling hexy using IK to calculate leg positions, but working entirely on the Servotor32. My aim is to connect a ps3 controller to hexy directly and control it as Michael does.

I've done the interesting task of solving the IK and doing the math, now I need to implement it properly which I would much rather grab someone else's solution if it's available.

Michael: I haven't copied your IK code, I have tried to implement my own based around some IK/Computer vision tutorials.
Yep that's right, I am taking the precomputed array approach but I think some of my point transformation functions are overly complex for what they need to do on hexy.

I have found http://blog.oscarliang.net/arduino-hexapod-robot/ tutorial very useful, and am considering just copying the code and making the necessary changes for Hexy.
klims
 
Posts: 31
Joined: Thu Aug 01, 2013 1:31 am

Re: (inverse)-kinematics

Postby rpcook » Fri Aug 09, 2013 8:57 pm

klims, PS3 controlled Hexy is an awesome goal, good luck with it.

I'd schemed this as one of my projects with Hexy but I've not had enough time to have a go yet. Few problems that I foresaw were:
  • I don't know what the communication protocol is from the PS3 controller (apart from Bluetooth), but I'd be very surprised if it's a low baud rate serial interface, which I think would preclude direct communication between the controller and the (current) Bluetooth module.
  • Given that the stock build of software in the Servotor32 makes the onboard memory about half full. Implementation of full-robot IK code might be either too large to fit onboard, or the processor speed would limit the speed of the IK code such that Hexy's speed would be limited.
  • A way that I saw to get around these would have been to use an intermediate device (I'd schemed an Android device), which has enough resources to form two Bluetooth links (one to PS3 controller, one to Hexy) and do the difficult IK maths. Someone's already posted a port of PoMoCo to Android, which might serve as a decent starting point.
Of course, these are just my foreboding, ominous thoughts, and they might all be solvable. Just thought that other peeps might benefit from sharing these ideas.

Would really like to see a project like this work, as demonstrating Hexy at schools etc, the PoMoCo interface is a bit of an accessibility block. Having a games controller operate the robot will really boost the engagement of people with Hexy. Good luck to you!
Rob
A record of my exploits and musings with Hexy and Sparki: http://robcook.eu, content available under a Creative Commons license.
rpcook
 
Posts: 121
Joined: Tue Sep 04, 2012 6:42 pm
Location: Luton, UK

PreviousNext

Return to Projects and Upgrades

Who is online

Users browsing this forum: No registered users and 2 guests

cron