Control Systems Group Robotics Lab

Hexapod Robot Project

Bruce B. Abbott, Ph.D.
Project Director


The goal of the hexapod robot project is to provide a physical test-bed for developing and testing control-system architectures conforming to the principles of Perceptual Control Theory (Powers, 1973). The hexapod design was selected for this project because hexapods do not require complex balancing mechanisms (thus simplifying design requirements), and because it will allow me to take advantage of Richard Kennaway's hexapod computer simulation's control mechanisms.

General Design requirements include the following: With these considerations in mind, I have spent the past week or so researching components. At present, the following arrangement looks attractive to me:

Initial Status

The first step is to investigate the options provided by various commercially-available controller boards designed specifically for the hobbiest robotics market. I have completed a web search that identified several promising boards. These generally come with on-board circuits for A/D conversion (to get analog sensory signals into digital format within the controller), I/O pins capable of driving the R/C servo motors, a communications bus to allow data to be exchanged between controller boards, sensors, and PC, and a local eeprom memory that can receive and hold instructions and data needed to implement local control algorithms. An important consideration is ease of programming. Some of these boards come with high-level compiler software that allow one to write the control routines in languages compatible with BASIC, Visual BASIC, C, or JAVA; these are compiled and the result sent to the local controller's eeprom where it can be executed without further need for the PC. I've also consulted with two of our Electrical Engineering Technology (EET) professors about which packages might be best for this application.

After giving the options careful consideration, I have decided to order an OOPic II Controller Starter Package ($69) and will be conducting tests with it to determine the OOPic II's suitability for use as a servo controller and as a higher-level controller. I already own a small number of R/C servos to use for these tests, but will order different ones specifially for the project if the tests prove successful. The OOPic is capable of driving up to 21 servos but timing and memory demands would probably be too extreme for our purposes with that many, so I will be investigating how well the controller can work with either three (one leg) or six (two legs).

Progress Reports

Here is where you will find periodic progress reports documenting what is being done.

10 April 2002

Thus far no money has been spent, but Bill Powers and I have been considering possible designs for a first-order control system, using an R/C servo, that would behave like the muscle-system in a real organism. It would include sensors for "muscle" length and force, and a rate-based damping circuit that eliminates the need for a physical dashpot. I breadboarded and tested an initial circuit for this that Bill designed, but investigation of the way the reference position of an R/C servo is set suggests that the circuit will have to be modified to comply with the servo's encoding method.

Meanwhile I've been receiving assistance from one of our Psychology students, Stephen O'Shaughnessy, who has a degree in EET and programs computers for a living. He's worked up a preliminary drawing of the servo/controller circuit arrangement in PowerPoint format and has begun to read B:CP. He's also looking into the problem of programming the communication between our on-board processors via the I2C bus.

Recently I've been spending time reading a couple of Rodney Brooks' books in which he describes his "subsumption" approach to robotics. This resembles PCT somewhat in that it is a bottom-up approach, but there are significant differences and it will be interesting to compare our design's abilities to those achieved by the subsumption-based machines.

2 July 2002

Quite a number of things have happened since the previous report on 10 April. These include the following:

14 August 2002

The second leg demo, described immediately above, was shown off at the 2002 CSG conference and Bill Powers and I did a bit of playing around with a Delphi program that operated the leg through the RS-232 interface to Bill's laptop.

If your browser supports embedded movies, you should see a short continuous-loop movie showing a complete cycle of the leg (otherwise you will see a still photo). When the leg comes down on the coin immediately below it, note how the LED at the back edge of the brass strip lights up. Sandwiched between the coin and brass strip is a small square of conductive rubber pad, whose electrical resistance decreases with increasing applied pressure. The material is being evaluated as a possible element for Thoat's force sensors.

Also seen in the picture are the OOPic II controller (behind the leg) and the Scott Edwards serial servo controller board (mounted on the leg stand), as well as the batteries powering the system.

Related Sites

Under construction, but this is the place to find links to other robotics labs doing similar work.