home  >  BARG index
Previous BARG formation
Next     BARG Newsletter Issue 2, Winter 1984

BARG Newsletter Issue 1, Autumn 1984
BRITISH AMATEUR ROBOTICS GROUP NEWSLETTER ISSUE 1 Autumn 1984 British Amateur Robotics Group 171 Worting Road Basingstoke HANTS RG22 6NR President Dr John Billingsley Committee David Buckley Alan Dibley Colin Freestone Richard Moyle
Editor's Bit Richard Moyle ZYLATRON a home-built Robot Richard Moyle Polaroid Ultrasonic Ranging System Richard Moyle Robot Intelligence Alan Dibley Competitions Richard Moyle Dibley Tests - Level 2 Alan Dibley Walking Machine Competition Richard Moyle All material in this Newsletter is subject to Copyright
Editor's Bit This is our first newsletter, and we hope you will find it useful and informative. We have included details of two new competitions in the tradition of Micromouse and Robot Ping Pong. One is to encourage development of practical machine intelligence, the other to get you all working on one of the harder nuts to crack - proper walking ability. We also have articles on Robot intelligence and sound ranging, and details of a domestic robot we met earlier this year. Perhaps before we get to the articles, we ought to introduce ourselves. We have been very fortunate in getting Dr John Billingsley as our President . John will need no introduction to anyone involved in Micromouse, a competition which has taken much of his time for many years. His books on interface techniques for the BBC and Commodore computers are essential reading for droidsmiths. Our Chairman is David Buckley, inventor of Zeaker, Prowler and many other machines, and author of innum- erable articles. Our Treasurer is Colin Freestone of Innovonics and through whose good offices we originally made contact with most of you. Alan Dibley is one of the most prolific of Mouse builders, with a very practical approach to the subject. His Mice would probably win prizes at Santa Pod. My own pedigree is nowhere near as distinguished as any of these, but we all share an intense enthusiasm for Robotics, an interest, we are happy to see, in which you all seem to join. Robotics appears to be developing in much the same way as early home computing, with dedicated amateurs showing the way long before big business knows where to begin. Each of us is inventing his own wheel {not to mention chassis) in different ways, and by encouraging the interplay of views and ideas we hope we can help everyone to get on more quickly. Obviously this relies on information coming back as much as going out, so let us have your contributions, even if only in the form of questions. Someone else may have the answer, if not it might stimulate further development. Ideas do not have to be of a technical nature - see Alan Dibley's article on surv- ival. Completely non-technical, it is nevertheless at the heart of Robot design. The newsletter is not intended to be a dry scientific discourse, neither do we want to go to the other extreme. Both technical and lighthearted material is needed. We even hope some of you might carry out pitched battles with each other through our pages. Finally a few words of caution. Do not believe, automatically, everything you read about building or designing Robots, here or elsewhere. Robots are new, and designs have not yet stood the test of time. Also, the machine you want to build will almost certainly be different from the one being described. Think about your own requirements, and use what you read or hear about as an aid to that design if it is appropriate. Do not be afraid to experiment. Eventually, practical standards and solutions will become clear, but now is not the time. Let's leave mass-production to the car manufacturers for the present. Richard Moyle
ZYLATRON a home-built Robot Richard Moyle At the International Personal Robot Convention in Albuquerque earlier this year, we met Michael Otis of Aberdeen, South Dakota. Michael's main hobby is astronomy, and he had become involved in remote control systems while designing and building his own observatory. After that he was hooked, and as he had this Timex (the computer, not the watch) doing nothing, he decided to build a Robot.... Almost a year later he turned up at the Convention with his prototype Zylatron. The machine stands about two feet tall and runs on a three wheel base with the drive wheels at the rear. Its specifications are impressive, with most of the high-level commands programmable in ZX BASIC with Forth as an optional extra. Programmes load from cassette, or can be entered directly through the ZX81 keyboard. He has even fitted sockets for an onboard TV, The computer/machine interface is handled by DOS (Droid Operating System), but he wouldn't give us any details. The rear wheels are independently driven by two large servo motors, powered by a heavy duty car battery, and controlled through opto-isolators. The battery also supplies the electronics through a regulated distribution panel. All power lines are filtered and fused and the voltage levels are monitored by the mpu. A socket is fitted at the back to connect a separate battery charger. Multiple diagnostic circuits were included, with LED status indicators As well as the usual mechanical bump sensors, Zylatron is fitted with a Polaroid ultrasonic ranger, which can locate obstacles at a range of up to 30 feet, and a photo-transistor visual sensor. It can also detect ordinary sound at up to 25 feet, and has a speech recognition circuit which almost works. An electronic smoke detec- tor is fitted, presumably so that the Robot can tell when its been plugged into the charger for long enough. A phoneme voice synthesiser allows a potentially unlimited vocabulary, though it wasn't working when we saw it. The Timex/ZX81 has a 16k RAM pack, though Michael intends to expand this to 64k. In all, the machine weighs about 301bs and is very impressive for twelve month's work by one person. We thought it showed up very favourably with the commercial machines on display, some of which were very primitive by comparison. Remind me to tell you about them sometime.
Polaroid Ultrasonic Ranging System - Richard Moyle The Polaroid Ranging System uses sound waves to detect objects up to 35 feet away and is capable of a resolution of 1%. There are two versions, one of which trans- mits 4 frequencies between 49 and 60 kHz simultaneously, to reduce the effects of topological frequency cancellation. The other uses a single 49.1kHz signal, and being simpler may be more reliable. The single frequency version is sufficient for use in personal robots, and it is this version which will be discussed here. The system consists of two main components, plus a suitable power supply. The Acoustic Transducer This is used for both sending the sound pulses and receiving returning echoes. The transducer is a disc 1.7 inches in diameter and is supplied with a perforated metal cover for protection. This gives it the proper mechanical insect-eye look which is de rigeur for Robots. A signal level of 300 volts is required to drive the trans- ducer, so it is essential that the connections to the circuit board are correctly made. Polaroid supply a short length of screened cable with slide connectors for this purpose. The alternative is to make solder connections to the back - a very delicate operation. The circuit should never be powered unless the transducer is connected. Polaroid supply two grades of transducer, Instrument and Commercial. Only the Instrument Grade should be used for ranging. They warn that the Commercial Grade is not guaranteed to work with the ranging system as it is liable to excessive ringing The Frequency Board The circuit board contains most of the electronics to control both transmission and reception, and consists of a digital drive circuit, a power transmit amplifier and a sensitive receive amplifier. Suitable control signals have to be supplied by the host electronics to initiate pulses. The circuit board will signal receipt of an echo, and it is the responsibility of the host to interpret this. The circuit clock is generated from a 435 kHz ceramic resonator and is sufficiently accurate to maintain a resolution of 1% {0.12 inches for distances under 10 feet). On receipt of a trigger pulse the board transmits for about 1 mS at 49.1 kHz after a delay of about 5mS, and then blanks the receiver for 1.6 mS. The circuit then waits for an echo for up to 62.5 mS. On receipt of an echo (or after 62.5 mS) the flag line will go high. The time between the pulse and echo is a function of the dist- ance to the object, and this can be calculated from the formula d = 331.4 * (k/273)*0.5 * t/2 where d is distance in metres k is air temperature in Kelvin t is time in seconds
At about 18 degrees C this simplifies to d = 171.1t, and taking the one-way distance as 171t gives an error of not more than 1% over a range of 12 to 24 degrees. This error is cumulative with the resolution error. The circuit cannot detect objects closer than about 11 inches as the return time is less than 1.6 mS. The answer for robot builders is to site the transducer a foot back from the front edge of the Robot. Objects further than 35 feet will not return an echo before the receiver times out. It is possible to modify the board to detect objects at distances up to 70 ft, with reduced accuracy, but the minimum distance also doubles. The amplifier automatically increases its sensitivity in 16 steps during the wait interval, as the signal strength of the returning echo is a million times less at 35 feet than at close range. This means the system is especially susceptible to noise interference at this stage, and it is recommended that as much other activity as possible is suspended during the transmit/receive period. Move- ment of the host system would in any case invalidate the measurement, unless some correction were made. The circuit board and transducer supplied by Polaroid as part of their Ultrasonic Designer's Kit are matched and set up to detect a 34 mm diameter sphere at a dist- ance of 1.3 metres from the transducer. A slightly different board is supplied when ordered separately, and some minor modifications are required by the purchaser. These involve cutting two circuit tracks and bridging others. Some components can also be altered to tweak the response, and full details are supplied with the board. It is unlikely that anyone able to interface the system to a Robot control system would have any difficulty. The Power Supply A system operating at 5 volts with a typical current requirement of 200 mA in receive mode and 40 mA in standby would seem to present few problems over the power supply. Unfortunately, in transmit mode the current requirement is 2.5 to 3 Amps in 1mS pulses at approx 70 mS intervals. This means that adequate provision has to be made to cater for this peak current and, in addition, suitable filtering must be provided to prevent pulses being sent back down the supply lines to disturb other circuitry. Polaroid suggest using their range of Lithium batteries, but for us this is likely to prove very expensive - as owners of the Sinclair Microvision TV set will testify. As well as the main components, additional circuits are needed to interface the system to the host machine. In the next Newsletter we shall give technical details, with practical suggestions on building a complete distance ranging sensor to incor- porate in your machine. We would welcome comments from anyone who has used the Polaroid system, or anything similar.
Robot Intelligence Alan Dibley Since the environment of industrial robots is unique to each application, I intend to limit the discussion of the meaning and definition of intelligence to robots in the real world, or a small part of it, the domestic environment. I propose that the attributes needed by a robot to succeed in this environment are those needed by any organism which makes a living on this planet (or any other, I expect). The are stated succinctly by the phrase "the selfish gene". That is to say, the driving force of the mechanism which motivates the organism is the survival of the organism, and of its offspring. The second aim is the dominant one (try removing a cub from a lioness, or even a chick from a hen), but for many situations is best achieved by striving for the first. Ferocity is only one of several strategies available to ensure survival of the specific genes of an individual; thus: 1 The lion Grrrr.., 2 The lion cub MUM! 3 The antelope Run like crazy 4 The armadillo A natural bank vault 5 The chameleon Where did he go 6 Plankton Safety in numbers 7 The rabbit Hide .... and so on .... Each kind of organism is equipped with systems which allow it to exploit the envir- onmental niche it occupies: extra good sight or hearing, long legs, good camouflage, Let us consider how these strategies might be adapted to improve performance in the domestic robot, and thus ensure its survival. But first, what tasks might it be given. I will discuss them under the headings Security, Clearing-up, Fetching-and- carrying and Other. Security This has most scope for the application of natural strategies. Consider the robot's problem when, at 3.00 am, it detects movement in the living room. A quick check reveals that the dog is asleep in his basket, the cat is asleep on the dog and all family beds are occupied. The lion strategy of just tearing out the throat of the "intruder" might not endear it to Auntie Alice if it was Uncle Arthur who was wandering around looking for the bathroom, unaware of the security system. I would suggest instead a symbiotic relationship with the dog. The procedure becomes : Actions Strategies 1 Go prod the dog The lion cub 2 Hide in the dog basket The rabbit 3 If the dog is not back in 60 seconds to push you out - SCREAM The lion cub again
It may not be a heroic scene, but survival is the objective. Detection of fire needs everything to happen at once, so a smart robot would grab the cordless telephone and dial 999 while shrieking at maximum volume and running like hell. These are all cowardly actions, but there are a lot more rabbits than lions. Think about it. Clearing up This is the sort of job that sells the robot to the rest of the family. A lot of special equipment is needed to allow the robot to behave sensibly in this task. For instance, at night while patrolling the house, the robot clears all the loose toys, dirty socks, muddy shoes and other clutter from the floor. It needs to differ- entiate between a pile of clothes and the family cat. The next-door-cat may, however, come under the heading of security - see under Security (prod the dog). Quite a clever sensing system is needed here if it is not to lose the co-operation of the dog. And under no circumstances should the dog be tidied up. Unidentified objects could best be dealt with by placing them on a table for reclaim in the morning, anything not reclaimed by the next night is automatically incin- erated. Could be a boon to a lot of harassed mothers. This sort of strategy is used by some kinds of fish which are immune from attack by predators because they clean parasites from the skin and gills of "customers" who come to known spots on the sea-floor to be preened. Fetching and Carrying These tasks are only of use to the humans in the house, but they had better be done properly, else it gets switched off. A fate worse than flat batteries. Examples: 1 Bring the morning post to the breakfast table, paying particular attention to perfumed envelopes, and buff-coloured, official-looking ones. 2 Bring the telephone to the nearest adult when it rings. 3 Come to the cry of "ROBOT" or "NAPOLEON" or whatever name applies this week. Complex instructions can then be entered on his keyboard. (Note, one of my prejudices is showing here). Other tasks Scope for the imagination here, for instance; 1 Walking the dog, (It's no good, you will have to get a dog, or at least a fierce hamster) 2 Answering the front door 3 Dealing with unwelcome callers by acting very unintelligently when answering the front door. An interesting strategy that needs further research.
It should now be clear that the key to success is sensor systems, and success can be equated to our meaning of intelligence. The effective gathering of data about the environment, and the processing of that data to prompt some action in aid of surv- ival is characteristic even of plankton. Robots should emulate living organisms in as many ways as ingenuity can conceive to ensure their propagation throughout the Earth, and tomorrow, the Universe. That could be a warning....

Competitions Most of you will have heard about the Micromouse Competition, and most of John Billingsley's more recent Robot Ping-Pong Contest. These have, or will have, the effect of encouraging development of particular Robot attributes such as real-world imaging, positional stability, vision and speed of response. Along the same lines, we are now introducing two further competitions, designed to stimulate further development. The first is an extension of Alan Dibley's article on Robot Intelligence, and is split into two levels. Level 1 is a preliminary stage, with separate sections for arm Robots and small mobiles. Arm robots are required to distinguish between diff- erently shaped objects, and place them into separate bins. Mobiles must distinguish between spheres and cubes, disposing of the former while not disturbing the latter. In Level 2, the Robot is required to deal with several real-world objects in approp- riate ways. Success at this level will take some time to achieve, being modelled more on the real world and requiring more versatility in the machine under test, Remember, though, that Micromouse seemed daunting at first. If the Dibley Test seems a little too tame, try our other competition. This is to build a machine capable of walking over a course strewn with various obstacles in the form of steps. The machine does not have to be a true robot at this stage, as directional motion may be controlled manually, but the mechanism of walking and climbing must be under computer control. The computer can either be on-board or remote, but the mobile part may not be more than 50 cm in any dimension. Formal rules and descriptions of both competitions follow, and additional copies are available from us on request. We expect to hold the first heats towards the middle of next year, depending on response. Please let us know as you develop your contes- tants, so we can arrange for appropriate facilities.
Dibley Tests - Level 1 These tests are designed to assess manipulation and shape recognition abilities. a)Arm-equipped Robots The test area is at least 50 on square, and is divided into three sections. The work area is the front half, and the Robot is placed here, in front of two trays. Each tray is one quarter of the test area. The one on the right holds six geometric objects of differing sizes and shapes. To the left is another tray with six comp- artments. Dividing walls are not more than 5 cm high. The objects are made of polystyrene foam and may be any mix of the following: Cone Sphere Cube Tetrahedron Cuboid Torus Cylinder Tube Pyramid They may vary in size from 2 to 10 cm, and may be of any colour. Initially, colour and texture will not be relevant, but at higher levels this could change. The objects would be selected at random immediately prior to the test. The Robot's task will be to sort the objects by shape, placing each kind into separate compartments, irrespective of size. An overall time limit of 15 minutes would apply, with any number of retries allowed within that limit. Success in completing the task would be of greater importance than a quick but inaccurate performance. b) Small mobiles The test area is a Robot Arena (approx 2.4 metres square) with 5 cm outside walls. The Arena is painted matt black and joins in the floor section will be covered with tape. Placed at random within the Arena will be 6 spheres of about table-tennis ball size (but not necessarily all the same size) and 6 cubes of similar dimensions. The Robot's task is to locate and dispose of the spheres over the side of the Arena, without moving the cubes. Again an overall time limit of 15 minutes would apply, with retries. Machines must be small and light enough to operate safely in a standard Robot Arena without damage to the Robot or Arena. Larger machines would be permitted to take the test, subject to their being able to reach all parts of the Arena from outside the walls, for example an Arm Robot with a reach of 1.5 metres at a height of 1 metre and able to move around the perimeter of the Arena could take part.
Dibley Tests - Level 2 This test is designed to assess shape recognition and judgement faculties in first generation domestic robots, with a Robot IQ requirement higher than that involved in Level 1 tests. The test area is a Robot Arena as used in Level 1b, which will be littered with the following items of domestic debris: 4 socks 4 shoes 4 tennis balls 4 fifty-pence pieces 4 wooden building-blocks 1 cuddly toy 1 clothes basket 1 four-legged stool The Robot's task is to locate and deal with each item as appropriate: Socks in the basket, shoes in the corner. Tennis balls thrown out of the arena. Money kept by the Robot as loot (until the end of the test, that is !). Blocks built as a tower - robots like to play too. The cuddly toy should be treated with respect in case it bites. Search under it if you like, but put it back, carefully. Avoid the stool. Use the basket for the socks. Time allowed is 15 minutes, with retries allowed within the limit. 4 points will be awarded for each item, with a further 20 given by the judges for panache or enthu- siasm. Maximum score 100. External power supplies by trailing lead will be permitted, but otherwise the Robot must be self-contained. Machines must be small and light enough to operate safely in a standard Robot Arena without damage to the Robot or Arena. Objects will be within their normal range of sizes, and will be distributed at random immediately prior to the test. The objects would be replaced in the same positions for any one series of tests.
Walking Machine Competition This competition is designed to stimulate the development of legs as a means of locomotion for small Robots. The test area is a standard Robot Arena (approx 2.4 metres square) with 5 cm outside walls. The Arena is painted matt black and joins in the floor section will be covered with tape. Placed within the Arena will be an unspecified number of obsta- cles consisting of wooden blocks with a maximum height of 5 cm. Obstacles may be placed on top of each other, but there will at least 10 cm between any two changes in height. A course will be laid out around and over the obstacles and will be the same for each competitor. The walking machine will follow the course, climbing over any obstacles, without leaving the marked route. The machine may be guided by a human Minder standing outside the Arena, by way of umbilical, radio, infra-red or even voice. The Minder may not, however, physically touch the machine once started. Guidance commands, however given, are limited to instructions of the nature of "Left", "Right", "Back", "Forward", "Up", and "Down". The machine must interpret any commands, or otherwise get around the course, with the mechanism for walking, climbing and descending completely under computer control. The computer may be on-board or remote. An overall time limit of 15 minutes will apply, with any number of retries allowed within that limit. The fastest machine to complete the course successfully will be the winner, alternatively, should no machine complete the course, the winner will be the one to get the furthest through the course. The mobile part of any Machine may not exceed 50 cm in any dimension (excluding any remote connections) and must have independently-controlled legs. Wheels masquer- ading as legs will be disqualified. Machines must be light enough to operate safely in a standard Robot Arena without damage to machine or Arena.
Next -
BARG Newsletter Issue 2, Winter 1984