Big Quadruped 'Dog'

[NicholasLee]

The big quadruped robot shown above was one that I built a shocking 12 years ago!!

Yesterday I got this out of storage, blew the dust off it, and took some better photos to show you.

In the next few postings I will try and illustrate how this project was constructed (and give pointers on how I would have designed it all differently given an extra 12 years of engineering experience .)

Regards,
Nicholas Lee

-------------------------------------------------------------------
EurIng Nicholas Lee BSc(Hons) CEng MIET MInstMC
Consultant Engineer

[Matt Denton]

Looks like a pretty complicated project.. 12 years old! does it still work?

[florek]

Nice one. Good construction for dynamic gait.. well.. almost That's called engineering Plenty of electronics.. was it walking only on and,or ttl gates?

[NicholasLee]

The quadruped interfaced to the pc via a 40-way ribbon cable to a digital IO card.
Windows software written in C++ then controlled the gait by sending sequences of joint position patterns (aka postures).

It was a bit like stop-motion animation. You would use a PC joystick to position the limbs and then store that posture information. By playing back a seqence of stored postures you could get it to go through any set of motions you wanted, including walking.

The robot still powers up ok, but the problem I have now is that the digital IO card uses the ISA bus, and they haven't made a PC with an ISA bus for years!

I'm going to have to find a more 21st century way of connecting my robot to a PC if I'm going to get it walking again.

Regards,
Nicholas Lee
-------------------------------------------------------------------
EurIng Nicholas Lee BSc(Hons) CEng MIET MInstMC
Consultant Engineer

[NicholasLee]

This is a top-down view of the electronics. All the instructions were sent as parallel data rather than serial connections, so there's rather a lot of ribbon cabling. The central PCB uses discrete logic chips to decode the address bus on the 40-way ribbon cable from the PC into 16 channels of 8-bit bi-directional data. 4 Channels of data go to and from each leg control PCB.

Originally the robot was to have 16 degrees of freedom with each leg PCB controlling 4 motors. This would have given the shoulder joints the ability to twist as well as move up-down, left-right.
However during construction it was found that the resulting robot was too heavy to support itself and so the mechanical design was simplified to a 12 DOF design.

Regards,
Nicholas Lee
-------------------------------------------------------------------
EurIng Nicholas Lee BSc(Hons) CEng MIET MInstMC
Consultant Engineer

[Matt Denton]

You should keep an eye out on ebay for a small industrial PC from Advantech. I picked up a great little unit with two half sized ISA slots, pentium mmx 266, 64Mb ram for about £40. I also have a similar system that will take four ISA full length cards. ISA is still used for industrial applications.

[NicholasLee]

Here is a picture of one of the leg PCBs. This uses a PIC16F84 programmed in C, which uses ADCs to measure the potentiometer voltages which indicate the joint angles.
A P-I-D loop servos the motors to the positions requested by the 8-bit data written via the ribbon cables.
A pair of L298N dual H-bridge FET drivers are driven by (20KHz) PWM waveforms from the PIC and this powers the motors. (Only 3 of the 4 motor outputs are used)
The small trim-pots shown at the bottom of the PCB control the calibration of the joint angles.
There are SMT discrete logic chips covering the underside of the PCB.

Regards,
Nicholas Lee
-------------------------------------------------------------------
EurIng Nicholas Lee BSc(Hons) CEng MIET MInstMC
Consultant Engineer

[NicholasLee]

This is a picture of the underside of the robot to illustrate the mechanics. (I unbolted the legs at the shoulder joints to make photography easier)

An extruded aluminium H-frame skeleton supports the four legs.
The four shoulder-joint motors are MFA-Como drills 919D motors fitted with 810:1 ratio gearboxes.
These give a huge amount of torque but move fairly slowly. The robot isn't going to get above a walking gate and break into a run.
The reason for this is that originally I used 148:1 ratio gearboxes, but the robot's wieight was so great that it couldn't stand up properly, so I swapped them out for the higher ratio gearboxes.

Using ligher materials, more compact electronics and motors with rare-earth magnets and planetary gearboxes would all have improved the power-to-weight ratio and allowed the use of lower gear ratios. This would in-turn have permitted faster joint motions and alowed experimentation with running and bounding gaits. However, rare-earth magnets and planetary gearboxes were beyond my budget so that was that!

Regards,
Nicholas Lee

-------------------------------------------------------------------
EurIng Nicholas Lee BSc(Hons) CEng MIET MInstMC
Consultant Engineer

[NicholasLee]

This is a picture of the underside of one of the leg units.


This is a picture showing the gearing of an attached leg.

These two pictures illustrate how the two motors attached to the skeleton are able to both move the leg forwards-backwards and side-to-side.
This is an unusual, (possibly novel) gearing arrangement, as normally one motor would have to rotate the first axis and the weight of the second motor. In this scheme the weight of the moving part of the leg is dramatically reduced.

Each motor's drive is taken through 90 degrees by mitre gears onto an axle that has a potentiometer on one end to measure the joint angle.
At the centre there is a differential gearing arrangement such that if both motors rotate together then the central 'spider gear' moves with the differential gears, swinging the leg side-to-side.
If the motors are rotated in opposite directions then the central 'spider gear' is rotated by the differential gears, but as the leg is attached at a right-angle to the 'spider gear' this rotation causes the leg to swing forwards-backwards.

Regards,
Nicholas Lee

-------------------------------------------------------------------
EurIng Nicholas Lee BSc(Hons) CEng MIET MInstMC
Consultant Engineer

[Matt Denton]

I spotted your use of the differential system in one of your earlier pictures.. very nice!