28 August 2017

K9 Diaries - 4 days to go

Dog tag
Its come down to the last 4 days now, and I've persuaded Debs to make me a collar, and I 3D printed a dog tag.

The construction was easy enough, she covered a strip of leather with some tartan fabric and added some Velcro to the edges to fasten it. 

The D ring that the tag hangs from was taken off a collar that once belonged to her dog Mitzie, who died 17 years ago now, Debs kept her collar and thought it fitting to use a part of her beloved dogs collar on my beloved doggie project.


The 3D printing in the tag didn't stand out as much as I would have liked, so Debs applied a couple of layers of 'Empire Gold' Pebeo Gilding Wax to the raised areas, she likes it when she gets to use her hobby on our robotic projects.

head, primer and more filling
Unfortunately with just 4 days to go to the event, K9's head isn't ready yet and I've not figured out how to attach it to the body and its becoming obvious  that more filling is required.

After drilling the holes for K9's ears, and the visor cut-out, I primered the top of the head.

Debs seems to be getting worried, and keeps asking me if it will be ready in time! Whats that all about?

After a bit of sanding it'll be on to the top coat.

So I nearly have a head, but it is time to be thinking of having a couple of days off work to give K9 a better chance of being ready :)

 

top primer


Now the control panel is painted, I can move on to assembling the components. The bezels I've printed fitted into the 2 large holes and the 12 indicator lamps took no time at all to fit. If there is time, I'll fit 4 momentary switch contacts to the top 4 switches, and use these as manual operators of the ears etc

Wiring the control panel is time consuming, and the software for the Arduino isn't started, let alone complete.
Fortunately it's not a difficult job as there are plenty of resources out there for Arduino, and the experience with programming the Chompalots eyes and walker robots servo sequencing has put me in good stead.

26 August 2017

K9 diaries - 7 days



Top of the head
7 days to go, and I've still not finished anything! I've still got a pile of unfinished dislocated bits. The head has at last been made and tidied up. I was scouting round trying to find something to use as a mould for the nose, but nothing was forthcoming so I modelled one on the CAD system and printed it on the 3D printer, then just stuck on with a bit of double sided tape to the head pattern before making the mould.

control panel mould

The control panel mould is a bit rough and ready but seems to have come out all right. I used copious amounts of wax mould release.

Below is the control panel and top of the head being produced in their moulds



The control panel is out of  its mould, and the Easylease release agent performed reasonably well,

Another package arrived today, hopefully the panel switch/lamps. The battery charger arrived yesterday,  so it is all coming together maybe.





Control panel, pre-tidy
I couldn't resist perching it on top of the body shell after I'd tidied it up a bit (below).

Now the control panel has appeared, the hole in the body shell can be made to suit it. Then the accommodation bits to stop it falling straight through need to be added. Then it is on to the body being  primed and painted, after a bit of sanding. This would be a massive step, when K9 starts  to come into view.




trial perch

Once the body is painted all the systems can be fitted and integrated. The sub assemblies collected together and we will see what problems and missing bits we still have to sort out.

The side panel is going to be a major part which needs making. There is no primer remaining and I think I need some more MDF!


My attempts at tagging K9, with primer at this stage






22 August 2017

K9 Diaries - 11 days to go

Head Mould

Only 11 days remaining before the appearance of k9 the wonder dog, so I'd better get my skates on. Fortunately I've been working slowly but steadily towards the ultimate goal. Sometimes it seems like he'll never be done, and so many different things to organise, and co-ordinate, with things slowly lumbering into position, it's like the aligning of the planets.


Fortunately the head mould did come apart mostly OK. Just a little sticking on the front of the head in the mouth area. Relatively cleanly, so only a little tidying up and it and it should be good to go.











The control panel is also progressing, this needs to be completed so I can gauge to correct size for the hole to accommodate in the body. Then I can move on to prepping and painting, before fitting all the systems in the body shell.
I decided on a two part mould. Hopefully the two parts will pull apart easily. Again with the experimental, I went for plain MDF with no primer, just 6 or so layers of wax release agent, all holes and imperfections filled with wax.

Both sides of the control panel mould are now covered.

16 August 2017

K9 Diaries - A Body in the Garage

Removing the mould
It began like any other evening, visiting family, what I really wanted to do was to get home to deal with the body in the garage. I had quietly secured it the previous evening, and I just needed time for the idea to gel. When I finally got into the garage, it was like driving a wedge between reality and a shadow. How would it turn out, would this be the end of a dream? Would the Fibreglass police turn up at my door and charge me with insufficient application of release agent?

One side revealed
After the wedges had been gingerly tapped in, the mould began to release again quite easily at the top, but it seemed to be sticking in the recesses again. Oh the torment, until with a cracking sound it just popped off. I was so relieved!

It also looked quite good for saying I'd taken a chisel to the mould.

Then it was on to the other side, I wasn't out of the woods yet.


Right side and front.

I was on a roll, the wax release agent is definitely is the way to go. Perhaps I'll try Easy Lease again in the future but I'll test it out on some not so important bits before I roll that dice again.

Here are three sides revealed, it was going too well!








Time for the bum


The rear section was a little  tricky, I thought it was going to give me some problems, but in the end just a little gentle teasing and plastic wedges tapped in all round, was all that was required.

The top section was a breeze. So this is it, K9 has a body shell, just a bit of a trim to remove the excess fibres around the edges and a little light sanding required around the edges.

K9's body revealed
Photos by my lovely wife, Debs :)

K9 Diaries - 3 Weeks to Whooverville 09 - PANIC MODE

Front and rear panels
After soaking the MDF and then scraping the primer off with a chisel, I've finally got a mould. There is still work to do cleaning it up, so I believe I need to fill any defects and then give the panels a light sanding.

This will probably take me through the weekend leaving less than 3 weeks to finish K9. Oh woe is me.

3 Weeks and counting - The situations is: Still no body, head, side panel or control panel.

The Motorised base is still good, and batteries have arrived, but charger needs sorting out.
Raspberry Pi is good, I2C connection to Arduino works.
Screen is good and the HiFi Berry should be great.

If I  can press on with the fibreglass this weekend, I may have a body shell and maybe a head.

The control panel could be fabricated quite quickly, and side panel is a quick job (famous last words).

So just a bit of filling where required, then onto waxing before fabricating the body shell......




It's a mould, huzzah. Not as pretty and shiny as it could have been but hopefully serviceable. With six coats of wax mould release agent, I'm pressing on!

Still no head, and I'm dreading trying to remove the bit's I've attached.

6 August 2017

K9 Diaries - Sticky Situation


With high hopes, I started gently inserting wedges and tapping them in. The mould should have just popped off.

After carefully working my way round tapping in wedges, it became apparent that it wasn't going to be that easy.

Unfortunately the release agent hasn't done its job. Rather than separating easily, I had to apply excessive force, which ripped the recess out of the pattern.

Wailey, Wailey!

I now had a big hole in the pattern and was quite anxious about the rest of the mould.








This is strange because the top half has released reasonably well, but in the worst possible places, the recesses, the mould has stuck to the pattern.

I messaged easy composites to get their take one the situation and was told that on new moulds they recommended 6 coats Doh!

So the issue was repeated on all four panels.




Side panel
The same with the other side, as well as front and rear.


In the end, I've had to demolish the entire K9 pattern to get the mould off.

This could slow things down a bit :(

And here's the remains of the K9 pattern, 'tis but a scratch. I'm sure that with a bit of glue and a polish we can get it back together and have another go.

The plan moving forward is to remove the MDF from the mould. Then it's the primer coat that'll need scraping off. Then clean up the surfaces.

No doubt a little filling and sanding here and there, but don't get too hung up about scratches in the surface just light sanding and smoothing.





After a day of soaking and scraping, the majority of the MDF is removed, but the primer is still attached. 

I've had some success scraping the primer off with a chisel, as long as I don't scratch  the mould surface too much it could be the way forward. Either way it's a lot more work, and time.

5 August 2017

K9 Diaries - Full Glass Jacket

Side prepared, ready...
The last side, it's been a long journey, but it is nearly over, It seems like I've been making this part forever, I've made a body, only to make a shell round it. So very soon I can move on to making the the actual body shell, that will be K9!

The final count down
I've received some more matting, this lot is a heavier weight so will make thicker layers, but will also require more resin.

Today, I've given the right hand side of the pattern a coating of  fibreglass, so that's all the sides done. I'm concerned that I may not have made the layers thick enough, but we will see how it goes. The thicker matting seemed to be easier to work with than the lighter weight matting I was using, so in future I'll use the 450g/square metre


Ready to break apart
After a bit of tidying, removing the hairy perimeter, we have a clean edge to see the interface between the mould pieces. I've drilled holes to hold the pieces together when it's being used as a mould.

Tomorrow is the big reveal, the unwrapping day, when I'll find out what sort of a job I've made of it.

Still if it's a dogs dinner I've still got time to have another go......maybe...So tomorrow will probably be more prepping for the big day.

3 August 2017

Digit Evo1 - One Giant Leap for Digit Evo 1

Digit at the Games
When we reverted back to Evo 1 Digit, I was pleasantly surprised to find that with just a little more work Digit was good enough to enter in to the event.

After sorting out the tables in the software that told the servos what position they had to be and when, I came to the conclusion that the chassis needed to be stiffened.

This idea coming about just a few days before the event, became a problem since it changed the centre of gravity again of Digit, which meant re-writing all the servo position tables.

However after after extensive fiddling and re-iteration of the code, the job was done.

Trip



Digit Evo 1 managed 3 metres in 27 seconds, on our kitchen floor, that time and distance equated to 2 minutes 42 seconds over the 18 metres, which we were reasonably happy with, so we packed everything up ready for the event.

So on the 8th July 2006 we went along to the event in Corby, Northampton where Digit Evo 1 weighed in at 700 grams, and passed the tech check with flying colours.

Leaving me free to check out the competitors, and watch the other competitions whilst we waited for our turn to take part.



Small Steps

Two of the competing robots 'Trip' and 'Small Steps', have overlapping feet which allows for a simple stepping movement to occur with no leaning required. However, these are quite small robots, with quite small steps, so they would need to have very fast steps to achieve a good time over 18 metres

Robbie and Gorgon are very sophisticated robots capable of achieving remarkable poses and stunts,  Their precision and sophistication meant that they were not optimised for racing, but they entertained, with their head stands and rolls.
Robbie and Gorgon



Barry
The robot 'Barry' was another robot with overlapping feet, but again, it had quite a restricted stride length, which explains why it took as long as it did to traverse the 18 metre track.

Road Runner 2
Digit's only serious competition was 'Roadrunner 2' which was an intriguing robot, with high speed feet. However during the race it struggled with direction and had to stop and realign which slowed it down considerably.

When Digits turn came, Digit got off to a flying start, but he started veering off course, I tried to steer him as best I could, this involved dragging one foot along the ground a little and is more of a directional bias than steering.

Digit rode up on the lane dividing strip, which actually helped me steer him straighter and recovered and so he put in a blistering time of 1 minute 46 seconds, despite having some steering problems.

Biped Race Results

We were thrilled to win the UK Robotic Games Biped Walker event with Digit Evo 1.

Receiving Digits 1st Place Certificate

Digit Evo 2 - A Step too Far

Digit Evo 2
Digit Evo 1, just wasn't fast enough or sophisticated enough for my liking, so I stored him away and began work on Digit Evo 2.

Evo 2 Knee
With Evo 2 I intended to give it proper legs with jointed 'knees', which would allow him to raise his feet off the ground, rather than leaning over to raise a foot clear.

However, the control of the servos relied on sequences of positions stored in tables in the micro-controller. The values were selected in real time for the servos, with no feedback. Feedback being necessary so that the robot knows how it is positioned within its environment and how it is walking. Many of us may recall walking up the stairs in the dark, and taking an extra step at the top of the stairs, where in actuality there is no extra step, and subsequently stumbling.

Our eyes normally tell us where the steps are, if being able to see the step is removed, we rely on memory and other senses. We also get feedback from the feel of our feet on the steps, Evo 2 has only the memory of where to place his feet. So if things start going wrong he doesn't realise and just carries on trying to walk, and has no ability to correct or accommodate things not being quite as he expects. All with a lot more servos, so considerably more complicated to get walking.

I did consider gyros and sensors etc, but at that time I didn't have the technology to proceed with that, and I was programming the micro controller in machine code, so it would have been a major leap in programming to accommodate the added complexity. 

Evo 2 Foot
I soldiered on trying to sequence the movements. Controlling the hip there is one servo for both legs striding, and one servo to twist one of the legs, for direction, then in each leg there is the  the knee servo and 2 ankle servos mounted on the foot.

Looking at the construction now, it does look a bit spindly, and the joints, although machined do not look like they would be stiff enough to precisely maintain the axial position, and would be a source of twisting etc.

Evo 2  Hips


The connecting rods are carbon fibre tubes 6mm in diameter and quite stiff. Another observation is that there is no shock absorption, so those poor servos had to put up with a lot of stress, and the servos range of motions look like they could have had a better arrangement.

Digit Evo 2 took a considerable length of time to build and wasn't as rigid as it could be. Instructing all the servos at the same time in a sequence, wasn't going to accurately position the legs and feet, as there is too much movement in the joints and twist in assemblies. So in the end all Evo 2 managed, was a quick stumbling fall every time.

I had a lot of fun working on him, but in the end we decided Evo 2 wasn't going to be finished in time for the 2006 UK Robotic Games, so we reverted back to Evo 1.

As a first attempt it wasn't too bad, but definitely needed development. However there wasn't another UK Robotic Games so poor Evo 2 has been stuck on the shelf along with all my other distractions.

Digit Evo 1 - One Small Step for Robot Kind

Digit Racing Against the Clock
In 2006, an event occurred called the UK Robotic Games, which had a biped walking robot category.

After about 2 nanoseconds of thought I, decided that I must enter as it was just too tempting to make a walking robot.

So after much sketching and head scratching I came up with a design and Debs christened it Digit, a name that stuck.

Digit was quite utilitarian, in that there was no attempt to make it animal like or human like, it was just built to race.

It is a statically stable biped, which relies on keeping its centre of gravity over it's feet at all times.
To do this it leans one way to raise one foot off the ground then moves that foot forward. It then proceeds to lean the other way, then raises its other foot off the ground and moves that forward etc.

Movement of the stride and the leaning are done by 2 servos. The control of the servos is by pulse width signals coming from a Microchip micro-controller.

To make the servo change position at the required rate, a sequence of positions are read from a table and translated into pulses at the correct time.

The correct table is selected depending upon a second micro-controller, which decodes the signals from a radio control receiver, such as forwards and the speed, as well as direction.

I programmed 3 speeds for digit, which had to be carefully tested, as stability becomes an issue when leaning side to side, because if you're not careful, it can throw the robot over because of the momentum.

Mechanically the robot has 2 large feet, legs are in 2 parts for rigidity, and to keep the foot parallel to the ground.

At the body the leg parts pivot front and rear with 1 degree of freedom, as there doesn't want to be any sideways motion at the body leg joint.

At the feet there are 2 degrees of freedom, in the form of hinge type joints, the sideways twist and a swivel. After getting Evo 1 to a walking state, I was inspired to evolve the design, and so set to work on Evo 2.