Build log day 11:
With the power system in place I could go back to working on limb control. I connected all the limb actuators to the pump, then set them to point 90 degrees away from the body. I then turned the whole thing around and set it down on the ends of the limbs. It could stand in this position without a problem and remained stable when I tried pushing it with moderate force. When I tried moving the limbs however I ran into a problem. The square base of each limb generates too much friction with the ground making the limb unable to move. I solved this by welding on some steel marbles at the ends of all the limbs. That enabled the limbs to move but I ended up with another problem; the contact point of the steel ball is too small which made the whole machine unstable, as soon as I tried moving a limb, the other three would start slipping and the whole thing would lose its balance and fall over. I spent the rest of the day trying to solve this by better control algorithms but it seems it’s not possible.

Build log day 12:
As I was working on the solution for the balance problem I realized I will be needing several more actuators which means a lot more wiring. In order to keep it neat I attached an aluminum cable channel running along the centerline at the backside of the main body. While I was working on this I came up with an idea. With the help of the appropriate additives I can make the hydraulic fluid multipurpose. One additive to enable it to transfer oxygen from the bellows to the energy converter, and one additive to make it conductive so it can carry current from the energy converter to the pump. It took some research to find exactly which additives to use and in what proportions, but after a few experiments I did eventually get it to work. This way I was able to remove both the power lines and the air tubes, lowering the overall complexity of the system. Of course, this also makes the hydraulic lines extra critical so I added some red pigment to the fluid to make it easier to discover leaks.

Build log day 13:
After taking care of the cables and upgrading the hydraulic fluid I could go back to working on the balance problem. After thinking about it for a while I came up with a solution that I think can work. I have created a kind of feet that I mounted on the ends of the limbs using ball joints. I then attached two hydraulic actuators to each foot so that I could angle it freely in two dimension. These feet have a large enough contact area to prevent slipping, but because they can be angled the limbs are still able to move. It’s then just a matter of controlling the actuators for the feet to maintain balance while the limbs are moving. To be able to fully concentrate on balance and limb movement, I have moved all the control for the pump, the bellows and the energy converter to a separate program running on an Arduino. I have mounted this Arduino directly at one end of the cable channel so that the power systems can function independently without a connection to my computer.

Build log day 14:
Today I have been testing the new limb control algorithm and trying to get the Man to move. Test results are as follows:
1. Static balancing test – pass
2. Dynamic balancing test 1- can resist moderate push forces in all directions.
3. Dynamic balance test 2 – will give way if push force is too great but return to balanced position as soon as force is removed
4. Single limb movement test – limb can move, Man will retain balance with one limb off the ground
5. Movement test – Failure! It can lift two limbs off the ground and retain balance but is unable to put the limbs back on the ground in any other position than the original one. It is simply lifting the limbs and putting them down again without actually moving forward. I stopped testing at this point.

I tried to get it to move using a different a gait than the one I had originally planned. I had hoped single limb movement could work as a backup but that did not work either. The only way I could get it to move forward was to lift both limbs of one limb pair at the same time. The main body would then sort of fall forward until the limbs hit the ground. It could then move the two other limbs forward until it was back to starting posture. While this technically counts as moving I don’t consider it a viable gait. I will have to figure something else out.

Build log day 15:
I have started to see some degradation in the performance of the power system. The problem seems to be that the multipurpose hydraulic fluid is somehow being contaminated by the energy conversion process which offsets the balance of the additives. I have installed a Liver series filter unit between the energy converter and the hydraulic pump which hopefully can mitigate this problem.

Build log day 16:
Yesterday I started a 24 hour test of the filter unit. I checked the results today, everything looks good. After a lot of thinking I came to the conclusion that the only way to get the Man to move is to add an additional joint to each limb roughly at the center point between main body and the balancing foot. This joint will be partially bent in the standard posture so when the limb is lifted off the ground it can extend before the foot is set on the ground again. With the main actuator and the middle actuator working in conjunction, the main body will be pushed forward as the as the extended middle join returns to the original (partially bent) position. I’ve spent most of the day working on this. To fit all the actuators I ended up nearly doubling the length of the limbs, so that the middle joint is where the balance foot used to be, with another section of square steel tube going down to the balance foot.

Build log day 17:
This morning I tested the new limb setup and immediately ran into a problem that I knew might come but had hoped to avoid. As the number of actuators on each limb increases, so does the weight of the whole construction and it’s gotten to the point where the weight is more than what the actuators can handle. The end result is that, as soon as I try to balance the Man on its limbs it collapses under its own weight. To solve this I will have to lighten the construction in some way. I’ve started working on it but don’t have a solution yet.

Build log day 18:
My first step in making the construction lighter has been to reduce weight of the main body frame. After some calculations I have realized it doesn’t need a full frame all around. Instead I have replaced the aluminum cable tray running down the back of the frame with a length of box section steel tube. This functions as the main chassis rail while doubling as a cable channel. I then attached two cross members to mount the limbs on, making the one for limb pair A wider than the other so as to allow for limb clearance. I then made a kind of cage from 6 mm steel wire to house the power supply system including hydraulic pump, bellows, energy converter and filter unit.

The second step was to lighten the limbs. I have completely removed all the hinges used on the limbs because the actuators are able to perform this function on their own. I have also replaced the square steel tube with aluminum. For the lower limb sections I had to use two parallel pieces of aluminum to ge the strength I need, but the weight is still less than that of a single piece of steel. I think these weight savings should be enough for now, if not I will have to look into some more advanced materials in the future.

Build log day 19:
I started the day with a simple static balance test to see if my weight reductions were enough. The Man was able to stand on four limbs without overexerting the actuators with the pump running at about 80 cycles per minute or roughly 40% of max capacity. This is good enough for me to continue working on movement, but I will try to lower the weight even more in the future in the hope of gaining increasing the spare capacity of the pump. With this success I set about getting the man to move.

It was harder than I had expected to coordinate all the actuators but I managed to do it. Since it had already passed a static balance test, I quickly moved to dynamic balance. With the middle joints partially bent, dynamic balance was good. The more I bent these joints i.e. moving the main body closer to the ground, the better the balance became. When I straightened the middle joints the balance became worse; with the limbs completely straight, even a light push to the main body could topple the Man. For the sake of balance then, a posture low to the ground would be optimal. However, the severely limits the range of motion of the limbs and will hinder the Man’s movement. After a bit of trial and error I settled at a 100 degree angle for the middle joint in the basic posture. Once I had this nailed down I could concentrate on limb movement. It took the rest of the to work out a control scheme for the individual limbs.

Build log day 20:
I started the day by testing out the limb control control scheme I created yesterday and coordinating all four limbs. It was a success; I got the Man to move forward on the first try. I ran a few more tests and it worked well for moving forward and backward. Unfortunately it couldn’t turn or move sideways. I quickly realized what the problem was: except for the balance feet, the joints are restricted to one degree of freedom. I spent the afternoon installing extra actuators on the remaining joints so they are able to turn in two directions. Unfortunately I haven’t had time to test it yet.

Pages: 1 2 3