First of all I would like to thank the community for all the great posts that are really of big help for starter robot builders like me and especially Zenta and Xan for the great code examples that really help to understand whats going on in a hexapod brain:-)
James is build around the SSC-32 Servo Controller, a LPC2478stk development board(ARM) and 18 HS-645 servos.
The mechanical design is inspired on the phoenix design.
firmware:
The hexapod has inverse kinematics for every leg, body rotation and balancing. Sin/cos calculations are done in tables so no floating points are needed.
The robot is currently running on a 6v 3700mah battery, this is also powering the servo controller logic and microcontroller. Currently am also experimenting with a sonar that updates the LCD to show what he detects.
Your hex, James looks great! What kind of processor are you using? And what represents the lines on the display? Is it free space according to what the ping sensor measures?
You’re telling that you’re using sin and cos tables. Are you also using a Boogtan2 table?
The processor is a LPC2478, it is an ARM based processor that runs on 57MHz. I decided to use that board because it comes with a ton of features including a 3-axis digital accelerometer with 11 bit accuracy.
The lines on the display show what the ping did measure. 1 pixel on the LCD is 2mm measured. The angle of the line give the direction of the head.
If the line is white it means the ping measured something within a range of 640mm. If the line is green he detected an obstacle.
In the movie you see only 3 lines because the head sweep step was very big so he took only a few samples.
The head is still a work in progress. Eventually I would like to make that the head can follow for example my hand so the spider can follow me around.
As for the boogtan2 table, I only have one table that contains 90 sin values (0-90°), based on those I am able to get my sin/asin or cos/acos.
I do not use tg/boogtan2 in my calculations so I have to check manually if the calculated angles are correct based on the signs of the 2 input lengths. This is still something I want to improve tho since I am not happy with the code I made for it.
Sounds like a good board! Got any link? You should be able to post them by now
Good idea! I’m also have plans to let BlackWido or Phoenix interact with the environment (people). I was thinking of add some vision. But I’ve got so much plans and so short on time
I’ve got a version where I use tables for sin/asin cos/acos. Those are pretty easy to build with 1/4 of circle indeed. But the arctan2 is a whole different function and takes 1/2 of the circle. This takes a lot of space… I don’t have a great solution for that yet…
I knew it! Although Jeroen is a very common name in the Netherlands you don’t see a lot of them over here
Welcome to the forum!
Hi Alan,
Yepp, sorry. “Boog” is the dutch name for Arc or inverse. BoogTan2 is ArcTan2 indeed. I keep mixing up the Dutch and English names
Cool board! Did it come with a stipped linux kernell or did you build a costom one? How can you programm it? Are you using the thouch screen or can you connect terminal? Or use something like putty?
It comes with a kernell for the board and an image that had support for:
Ethernet
RS232 (used for Linux console)
SDRAM
LCD framebuffer
The kernell and image can be loaded via an usb stick or flash card. The files are build on your pc from an archive you get when you buy the board.
You can communicate with the board via a terminal.
For the hexapod I do not use uclinux.
Instead I initialized the board myself using eclipse with a gnu arm toolchain. Currently the controller runs from flash. The LCD buffers are in sdram. For communicating with the board I use a usb tiny jtag probe. Through the probe the controller can be upgraded and debugged.
Currently in the hexapod, 1 gaitloop takes about 22ms, this includes inverse kinematics for the legs, body rotation, body balancing, ps2 readback and translation from the angles into usable serial commands for the servo controller.
With the sonar included the time becomes ofc alot longer and variable since its dependant on the measured length.
I need to take some pictures still from his current state, will post em tommorow I hope.
You are developing a great robot! I’m particularly impressed by the processor.
Have you looked at using interrupts to handle this part of the code? For example you could send the ping and start a timer. Then use an external interrupt (when the ping returns) to trigger a part of code to check how much time has passed. In-between the processor can calculate other things (like IK).
The ps2 scan is pretty fast as long as you can keep the controller in analog mode. If you have to re-enter analog mode every time again it takes alot longer.
C
I did not consider external interrupts yet, gonna check that out. Sooooooo many things to check
.
