Hi, I have been looking for a mobile platform for the robot project I have been working on, and have long been looking at using wheels to move it around. Lately however I have become more interested in a
4-legged architecture as that would enable to robot to more easily get over door lists and similar obstacles.
I found some at Lynxmotion:
But these are too small for my project. I need something that could carry around 10 kg total weight (or more) as it is to be a robot based on a Mini-ITX system powered by a 8 Amp SLA battery (weights around
2.5 kg). What kind of servos would be needed to lift such a weight? I guess the weight would be supported by at least two legs, but I think I need to build it so that one leg could take most of weight by itself (in case it is tipping over).
I thank you for any information regarding this. JC
I am playing with something similar -> four legged machine, mini-itx (Epia-M motherboard, Eden processor, 40 GB 5400 rpm hdd ...) as brain.
There are zillion problems to say the least, for example, weight and numbers of servo motors (apsolut in kg, relative in Nm/kg), weight and numbers of batteries (capacity in Ah, relative capacity Ah/kg) and such.
My design is 3 DOF per leg, four legs, 3 DOF for 'neck', and 1 DOF for 'mouth'.
Servos are windshield motors, two types: first one is from 'Zastava 101' car (very often where I live), weights 1.375 kg with 13 Nm of torquee at 12 V from car battery (stall current ~ 8 A) and 24 - 25 Nm at 24 V (two bateries - stall current > 15A) and relative torque of 9.4 Nm/kg; second is windshield motor from Volkswagen Golf II, weights 1.25 kg with torque of ~
9.5 Nm at 12 V from car battery (7.6 Nm/kg). This second is relatively much weaker, but it is also very compact. Both of them are rotating at ~ 1 rpm at
For 'neck' and 'mouth' I use DC motors from battery operated drills (did't measure torque nor weight, but are with gearing much less then kilo; each of course. they usualy operate at 16.8 V, so running them between 0 - 24 V is not a problem).
Batteries are 12 V, 18 Ah lead-acid (not gel cell) used in motor bikes and weights 6 kilos. Each. Two of them are connected in series for driving servos at maximum 24 V. One battery is used for electronics and comp.
Body without legs is almost 40 kilos (almost 90 pounds) and ~ 55 kilos (~
120 pounds) overall. Legs are short, because I don't want it to have problems with standing up, walking around home (strait line) - uper DOF is
12.5 cm and lower is 15 cm long (but it is faster than aibo he he). Never tried more than one step (which was not automatically - I controled it almost entirely manualy).
I use printer port in EPP mode to control (to be correct - to try to control) it. EPP is great thing because one can send data/address on 8 bit bus. I don't like ECP, but one can use EPP+ECP port just as pure EPP, if configured corectly.
For example, first send address n. Than send data to that address (data that means speed for servo for example), than send addres n+1 and then data to that address (in my case it means direction of rotation, full speed, full stop and few control lines). If you decide that 4 bit speed regulation is enough, than you can control each servo with just one address & data write operation. Good thing about EPP is that data/address can be read, too. Control logic for EPP is simple and can be found around (viva Google).
hope this helps
It is still a total mess, and it will take some time untill it is ready for first 'public' photos ...
P.S. Mini itx is ok thing. If only it has one more PCI slot, it would be a killer ... ok, if it had second PCI, people would probably ask for third one :o)))
This is a difficult problem - and animals do it so easily. 4 legged architectures typically require dynamic balancing for most gaits, although you can get by with static [ie, 3-legged] balancing when using a creep gait [ie, lifting 1 leg at a time, like a cat creeping]. Building a quadruped to lift more than a few kgs is a real challenge. There are a lot of quadrupeds on the internet, and I have a bunch of links here:
Check out especially the walking machine catalog and Kimura Lab.
I've done quite a bit of research on this myself. The ideal solution is a hydraulic or pneumatic based system. Of course, these are often too complicated and costly for most hobbyists, so most opt for servos.
However, pneumatics are almost within our reach. A system composed of air muscles and nitinol actuated air valves would provide immense strength and flexibility, yet with relatively low cost comparable to high-end servos. Air muscles are simple enough to construct yourself. Plus, there would be no cost involved in maintaining a fixed state, as nitinol valves close when powered down. Still, it's hard to beat the encapsulated simplicity and cheap cost of the standard servo.
Unfortunately, for the lifting capacity you're looking for, adequately strong servos aren't going to come cheap, at between $40 and $60 a piece. Be sure to get metal geared servos, as anything less will likely strip under extreme weight. I recommend the Hitec brand as they're less expensive than Futaba yet just as reliable. Buying second hand, like from Ebay, can also save some money. As Mateo suggested, a realistic option is to simply construct your own servos. Lynxmotion has some photos showing how they interfaced standard servo control electronics to one of their 12V gear-head motors
I also recommend you ditch the SLA battery. The prime consideration on any mobile platform, especially legged, is weight. Lead batteries are just about the heaviest and least efficient battery you can use. Lithium-Ion is of course the ideal, but NiMH are cheaper, more widely available, easier to charge, and still far lighter than a comparable lead-acid battery.
Another aspect you should consider is your bot's sensory perception, especially important on a legged platform. The obvious necessities would be contact sensors on the feet, preferably variable, so your bot knows how much weight it's carrying. A servo's internal potentiometer can be center-tapped for external use, providing a simple angle sensor for each DOF. Also important is some combination of accelerometers and gyroscopes to detect linear and rotational motion, essential if you want to bot to be able to move dynamically.
You're on the right track using a mini-itx. It annoys me when people try to do projects like these with some underpowered PIC or BASIC Stamp. Some people have recommended the PC-104 boards, but I've found them either ridiculously overpriced or underpowered (usually both), especially when compared to the itx boards. Personally, I'm eager for the release of the nano-itx.
If you can find coreless motors with graphite brushes, you will get more wattage out of them. Automotive windshield wipers have only one adgantage, in that they are sometimes worm driven, otherwise they are terribly inefficient for their weight.
Lithium Polymer is superior to lithium Ion. Lithium Ion can discharge at 2C, whereas Li Poly can do 6C or 7C.
Consider quarter-scale servos, such as the GWS S666 or new Hitec HS-755HB (if you can get them). Cost is about $30 each depending on model, so get a bank loan first. Seriously, they can be pricey, but they are still less expensive, and generally easy to work with, than DC gear motors with external feedback.
As another poster mentioned, watch the weight. Actually, it's not the weight itself per se (within limits) but weight distribution. You want to avoid putting a lot of weight into each leg. Consider various reliable but simple linkage mechanisms that allow you to put at least one of the servos on the body of the robot.
Forget SLA batteries, in this day and age of modern battery chemistries. Mike suggested lithium polymer. More expensive, but with the right charger, you'll grow to appreciate it. (I do recommend just purchasing a charger and not trying to build on.) Polymer batteries can also be made to most any physical dimension, including the common "flat pack." Very handy for walking robots, where keeping the COG low is important.