I tried to take on board some coments about how most commercial robots run manically across the table to try and reach the end before they fall over.
I now know why…
The bot is tall and skinny with a very high centre of gravity. To balance on one foot long enough to move the other foot forward means an incredible angle is needed on the ankle. 9 degrees in this case. Moving from foot to foot means the robot sways and is on the verge of falling over, 11 degrees on the ankle is all it takes.
The weight transfer also causes flexing in the ankle which is enough to take it over the 11 degrees so I have to land the foot at 5 degrees then lower to 9 degrees to absorb the sway.
I think this is why most walks start with the robot crouching down to lower the centre of gravity (takes 13 or 14 degrees before it topples then)
Still its an early attempt. I may find a better way (turn it into a hexapod for example).
Any comments or ideas on how to improve it would be welcome.
I had actually bought all the brackets to build a BRAT myself, but then decided to design and build something custom. That’s Mega Scout and its smaller sibling, The Bipod came about. I will be able to use almost all of what I build as the Bipod to build Mega Scout - the same first four joints and feet for sure.
Stunning to be precise. Using the Old controller I had no end of problems but with the SSC it can walk out of my office, across the kitchen and harass the dog quite nicely.
Some 30 odd feet around electrical gear and not a glitch…
In the first video, it doesn’t look as if there is any shifting at the hip joint. When it takes a step, the upper torso seems to be on an angle rather than verticle. Perhaps put more compensation in the hip joints keeping the torso in a more verticle angle when taking steps?
I just looked at the second video and It’s amazing how natural looking the turn step is. That is perfect!
The 2HV is the older one, the 1HV is the one released Jan this year.
The 2HV didnt have hip rotation. The 1HV has ICS red servos for all but the shoulders which are 270deg 4024s
SN96. I did try using th hips to give it a more natural gait but the rotation throws the centre of gravity over to far. I deliberately kept the slight forward lean because with the torso upright it always seemed that either the toe or heel of the foot caught as it moved forward. I think I may add an additional position before and after the foot movement forward to raise the foot slightly higher first.
So far I’ve had two methods of learning, Trial and Error (more of the latter) and By Example. Thats been the most effective.
Trouble is there are few examples of good walks for a biped with rotating hips so I ended up watching RobotDudes strut mpegs over and over to try and see the intermediate steps. I also read derekzhans(sorry if thats wrong) thread from end to end to try and get hints and picked out what would work for my gangly, overly tall bot(good job he’s thick skinned) It gets me to the next stage but its got to be better than that.
I think you are right though. Its something to do with a hip movement thats lacking. More for aesthetics than anything really as walking is walking. I would love to got the route of derek but sensors and IK on the fly is a bit beyond yet at the moment(possibly always will be)
Accelerometers and tilts will help with the falling over bits.
Look at some of the old kinematic study films, or some of the more recent computer animations, showing a skeleton walking. The hips rotate with each step, but the waist also rotates slightly in the opposite direction, resulting in the pelvis tracing out a sort of figure-eight motion. This is isolated from the rest of the body by the lower spine, so the torso stays mostly still, other than the shoulders. With three axes of rotation - two hips and one waist - you could approximate the pelvic motion, though your torso would still be rigidly tied to the vertical axis running through the center of the waist rotation mechanism. In order to fully isolate the pelvic motion from the torso, you’d need two further axes directly on top of the waist rotation, and work out some fairly complex motions to null out the twisting and swinging imparted by the stride.
So far as existing biped gaits using hip rotation: In most of the examples of rotating hips I’ve seen on biped robots, the hips are only rotated in order to turn the entire system, and locked forward during straight walking. Turning is usually accomplished like a half-step, stopping or slowing when it’s balanced on one foot. At this point, the hip is rotated (or most often, both hips rotated in opposite directions), so that when the raised foot comes down, it’s facing in a new direction, but with the torso only facing about halfway to the new heading. Then when the other foot is raised, the hips are straightened, turning the entire bot the rest of the way to the new heading. This is an improvement over the common “sliding foot” type of turns that you often see on bipeds, since by rotating the hips, you don’t need to move the foot in relation to the ground during the turn. It can also be used, in conjunction with the ankle axes, to shuffle side to side using the toe and heel edges of the footpads.
In most cases I have seen, both hips rotate the same amount in opposite directions, something that can be achieved mechanically, though they usually have individual axes powering each hip rotation, so that they could rotate in the same direction if they wanted to.
Edit to add: Though of course, the most practical benefit to having a floating pelvic assembly on a bipedal robot would be that it would impart the ability to properly perform the Time Warp dance.
If you look at the youtube posts above you’ll see thats how I use the hips to turn.
At the top of the hips there is the rotation servo but also a second servo below that allowing lateral movement (the splits). The problem with thm is that until the sideways hip movement on robotdudes bot which has the rotation in a good mechanical postion, these are offset to the right and left so whatever power the servo has is lost in the negative mechanical advantage of the joint arrangement (if that makes sense)
Good work!
Sorry. I thought 2HV is the newest one because the KHR-2 came after -1 
my bad ^^;;