Walker Robot ver.1

This circuit is pretty simple (no reverse, no turning, only 2 chips), so it was just what I needed for this project. A number of other folks had touted the use of 74*245 chips for motor drivers, so I changed Wilf's circuit in that direction. It turns out this (like many things in life) had been done before -- here is a fairly "traditional" circuit layout:

While this layout is simple to build on a breadboard, it's a mess if you build your own PCBs, as I do (a number of jumpers are required). I modified the traditional layout to come up with a simpler PCB design (look, Ma, no jumpers!):download

Essentially all I did was to re-route the path that the quadcore process takes through the 74*14 chip -- rather than running down one side of the chip, then down the other, in my layout it "zig-zags" from side to side. I wound up building this circuit in a ".1" and ".2" variant; the ".2" version has provisions for (i.e., sockets for) filtering capacitors on chip power lines, the ".1" doesn't. Note that you can avoid the clutter of these capacitors if you buy IC sockets with power filtering capacitors built in (tho' these sockets cost about $2 US each, vs. a few cents for "plain" sockets).
Here's the life-size PCB "artwork" for the Walker 2:14:1.2 circuit. A higher-res (x4) version is available here.
I make the PCB via the "Press-N-Peel" method, and can repeatably produce these cards without much trouble.

Here are pictures of the top and bottom of a populated Walker 2:14:1.1 PCB, on a 1/4" grid.

Here's how to lay out the components on a Walker 2:14:1.2 PCB. download

Here, the D1 diodes are small (optional) indicator LEDs, D2 diodes are small-signal diodes (1N4148, or equiv.), C2 are (optional) filter capacitors (I use 0.22 uF), C1 are the quadcore timing capacitors, R1 are the quadcore timing resistors, and R2 are current-limiting resistors (1 KOhms) for the LEDs. Note that the filter capacitors (C2) always face consistent polarity power, so you can use polarized capacitors for them if you like; the timing capacitors (C1) see varying polarities as a normal part of quadcore function, and so must be monolithic non-polar caps.
The values for C1 and R1 determine the neuron firing times, and so are functions of the gear motors you use and the voltage you're running this circuit at. I use Mac floppy eject motors (which are efficient, if a bit slow) and run the circuit at 5V, and find that C1 = 0.22 uF and R1 = 2.2 MOhm works well for me. Your mileage may vary.
Note that the 74*14 and 74*245 ICs are "socketed," both to protect the ICs during soldering, and to allow for experimentation with IC subfamily. Similarly, the timing resistors and filter capacitors are plugged into individual sockets to allow for tinkering.
As is mentioned elsewhere, the 8-pin DIP socket allows a variety of "brain" cards to be swapped on and off a given walker.
For more information...

You can think of this circuit in two parts -- quadcore and driver. The 74*14 quadcore circuit is explained here; while the 74*245 driver circuit is explained here.
Andrew Miller used to have a site up with a walker tutorial; his tutorial was adopted by Bram Van Zolen here when Andrew shut down his site. Andrew's tutorial is educational, but pretty font-intensive. Less "busy" versions are here and here. Note that Andrew's tutorial is based around an earlier version of a 74*14 quadcore, with 74*245 driver (essentially an older variant of the Walker 2:14:1 circuit, with manual PNC circuitry).

WalkerRobot1 Schematic