Bing 3

[DerekZahn]

Building the next version of my robot Bing took a long detour; I decided to build myself a new mill and the first few attempts were unsatisfactory. However, it is now working adequately so I better get back to work if I hope to be ready for RoboGames! Here's a shot of the mill for the curious:



bigger: http://happyrobots.com/newmill.jpg

The build has started with the feet. At the very bottom, some rubber-coated thin carbon fiber pads:



bigger: http://happyrobots.com/bing3a.jpg

The feet will be hollow, holding batteries and some small electronics, so here's part of the foot:



bigger: http://happyrobots.com/bing3b.jpg

I really want to see if I can use sensor feedback to improve performance, so each foot will have pressure sensors to hopefully help with balance and so on:



bigger: http://happyrobots.com/bing3c.jpg

I'll continue to post reports as I make progress over the next couple months.

[Joe]

Nice work, Derek. That mill is a beauty, and it seems like Bing3 is off to a good start.

I noticed that your feet are basically rectangular. I started off with feet of the same shape on my 4DOF biped, but found that it was stepping on its own toes and heel a lot. So I built a new pair of feet with the corners taken off (basically a stretched octagon), and these work much better. The robot is able to make very tight turns without stepping on itself, and I haven't noticed any detrimental effect on its stability.

I think those pressure sensors are going to be really useful — much of what we do when we balance on one foot (for example) is equalizing the load on the foot. It's something I want to try someday, too — can't wait to see how it turns out for you.

Thanks for the post, and keep 'em coming!

Best,
— Joe

[DerekZahn]

I saw the shape of your bot's feet, it's a great idea. I need the whole rectangle for the battery but I might steal that for a future version.

Your nifty little bot walks better than any version of Bing ever did!

[DerekZahn]

Since I had a bunch of CAD files sitting around for months, I'm able to get right to work making parts. Here's a shot of a part as I work on it:



bigger: http://happyrobots.com/bing3d.jpg

The problem with CAD is that it's tempting to make things kind of complicated and then you have to actually build it! Still, progress is being made. Here's some finished parts:



bigger: http://happyrobots.com/bing3e.jpg

[DerekZahn]

Speaking of CAD, here's a cad sketch that I'm working from for the legs, so you can see roughly where I'm headed:



bigger: http://happyrobots.com/bing3f.gif

[Joe]

Looks neat, Derek. Can't wait to see how it progress. I'm envious of your milling machine — that looks like a real enabler!

What sort of servos will you be using? In your leg sketch, I think I understand the knees (two servos stacked together, for double the speed/strength), but the ankles and hips puzzle me. The way the brackets attach, I don't see where the servos are going to go without overlapping. Do you have a sketch that shows how the servos fit in?

Best,
— Joe

[DerekZahn]

As for servos, I've been collecting these for the last year and a half:



bigger: http://happyrobots.com/bing3g.jpg

Some of the joints will be using less expensive servos, I have some left over from the last Bing I can use.

Sketch showing how the two servos are aranged in the joints:



bigger: http://happyrobots.com/bing3h.gif

[DerekZahn]

Progress continues... Preliminary assemblies for leg pieces are ready (but the wiring and electronics still needs to be worked out):



bigger: http://happyrobots.com/bing3i.jpg

[DerekZahn]

Bing 3's feet are hollow. Each one contains a battery cell and some electronics:



bigger: http://happyrobots.com/bing3j.jpg

The purpose of the electronics is to turn the foot pressure sensors into a voltage that an A/D converter can read. These sensors (which you can get from www.phidgets.com ) have a resistance that is inversely proportional to the pressure. However, the resistance decreases exponentially with pressure, rather than linearly, which makes the response kind of painful. I use an op-amp to linearize it, making the resulting voltage (hopefully) more friendly for the future software that will use it.

[Joe]

Looks neat. I'd appreciate more detail (maybe a schematic and an explanation of how it works?) on that linearizing circuit. I can imagine that coming in handy for all sorts of sensors.

A question about the batteries: I see an awful lot of screw holes around the outside of this foot. Are you going to have to remove the sole to recharge or swap out the batteries? If so, any idea how long that will take?

(This is an issue I'm wrestling with on my current bot, where the batteries are rather hard to get to... but I'm reluctant to just mount them on the outside, either, so I'm trying to come up with some clever way to hide them yet keep them easily accessible.)

Best,
— Joe

[DerekZahn]

Sure, here's a simple diagram:



This is a "noninverting amplifier". The equation for this is:

Vout = Vin * (1 + (R1/R2))

In this case, I supply a constant voltage (say 1 volt) as Vin, and pick some appropriate value for R1 depending on the response of the sensor (which is R2). In this case I picked 3.3k. So the equation becomes:

Vout = 1 + 3.3k/R2

Since R2 (the sensor) is in the denominator, the amplifier linearize it (more or less). At no pressure, the resistance is infinite, giving an output of 1 volt. At R2 = 3.3k, Vout is 2 volts. At R2 = 1.7k, Vout is 3 volts. At R2 = 0.8k, vout is 4 volts, and so on.

There will be connectors outside the foot for recharging, though changing the battery will not be a quick task. I don't intend to change it.

[DerekZahn]

My favorite microcontroller board lately is the tini2131 from www.newmicros.com

At $29 it's a pretty good bargain for a 32-bit ARM chip running up to 60 mhz, 8 channels of A/D, 2-3 channels of PWM, and so on. Plus, it's nice and small.

To help Bing keep its balance, I am building a little sensor pod containing one of those processors, plus a pair of IDG300 2-axis gyro boards from www.sparkfun.com and a Hitachi H48C three-axis accelerometer from www.parallax.com. The idea is to package that all in a little box.



bigger: http://happyrobots.com/bing3k.jpg

It all becomes self-contained by screwing the top on:



bigger: http://happyrobots.com/bing3l.jpg

In general I'm saving the software issues until I can get Bing finished mechanically, but I did enough of the software for this to know that I'm reading the sensors properly and it is wired up okay.

[Joe]

[DerekZahn]

That baby orangutan is a nice looking little board, pretty similar to the MEGAbitty.

I expect to do a fair bit of processing on the sensor signals -- averaging, perhaps kalman filtering.

I don't think there's any best chip or board (if there were, the competitors would be out of business I suppose). It's a matter of taste.
[/url]

[DerekZahn]

One of the goals for Bing 3 is to work out some sort of scheme to keep the wiring under control. What I decided to try is dedicating a microcontroller for each set of 3 servos. Those microcontrollers generate the PWM to set the servo positions, and gather sensor data, and communicate on some sort of serial bus with each other. To help organize and process the data, each little tini2131 board gets a little daughterboard attached:



bigger: http://happyrobots.com/bing3m.jpg

Here's what it looks like after partial assembly. Unfortunately, the designers of the tini2131 decided not to route all three PWM signals to the header, so I have to solder a wire directly onto one of the pins of the processor chip, which is a delicate operation! Notice the posts (which are just chunks of wire cut off of a resistor) that will attach the boards together.



bigger: http://happyrobots.com/bing3n.jpg

Here is the result fully assembled. The power leads daisy-chain into the big connector on the left and then go out through the big wires. On top are headers for servo power, control, and feedback.



bigger: http://happyrobots.com/bing3o.jpg

Here's the board inserted into its home in one of the leg pods:



bigger: http://happyrobots.com/bing3p.jpg

Underneath, the headers are exposed for wiring:



bigger: http://happyrobots.com/bing3q.jpg

And here's what it looks like after it's wired up. The two yellow wires coming off the bottom are only needed for programming the microcontroller and will be removed before assembly.



bigger: http://happyrobots.com/bing3r.jpg

My favorite Robo-One robot is OmniZero and I'm borrowing ideas like mad from it. One of the neatest things about it is how the wiring is completely invisible and I'm going to try to accomplish the same thing myself. So there will be a cover of some kind on the servo pod to hide the final wiring, and you can see the wires coming out through the holes in the side of the pod. Those will go through a bearing and a hollow bushing, after which they can be routed to their destination. Because the wiring goes directly through the axis of rotation, the wires don't need to be loose and can be easily covered up. It's extra work to do it this way but the results should be worth it. The series of robots leading up to OmniZero have been doing this for a long time.