Need More Holding Torque through clutch/brake

[davidhere40]

I'm building a very long robotic arm: around 3 feet long. Maybe even 4 feet, if possible. I have AX-18A servos which can hold 18kg at 1cm. But, at 3 feet, that would reduce to only 0.14 kg. Also, the servos will quickly overheat if I hold it there anyway. Once the arm is outstretched (if that's how far it has to go), I need to be able to grab and lift a 0.3kg object and bring it back using the arm.

So, I need to keep my costs down and avoid overheating. So, I can't buy more powerful servos. What I'm thinking is to create a simple clutch that will lock a servo's joint when it has reached a goal position. This way, I can start out with the arm contracted, move the first joint, lock it, move the second joint, lock it, etc. This will dramatically reduce the current that the system uses and also the torque required of the servos.

But, I need to keep the clutch cheap and simple! It should be simple for reliability, repairs and cost.

Any ideas on how I can achieve this? I was thinking of using a cheap servo to move a screw in and out of a locking holes along the joint. But, not sure that's "simple". Also, it's either locked or unlocked. It would be nice if it could act as a "brake" also to let gravity do some of the work for moving the arm. That way I could lower the arm more gently to a desired position.  I was even thinking of adding cheap servos to the same joint as the AX-18A's. This might allow me to use the AX-18A's position detection and load detection, but get more torque for less money.

any ideas?

[newInRobotics]

Adding more servos and clutch systems will only increase weight, power consumption and complexity of Your robot.

Instead of servos I would go for stepper motors. Servos are very "wobbly" where steppers are not. Or, You can have normal DC/stepper motor and worm gear and pinion setups for joints that are required to lift heavier loads. Worm gear and pinion mechanism is self locking and is fairly easy to implement.

[Soeren]

Hi,

Just one more vote for worm gears in that application.

Here's some inspiration - scale up or down as needed:
http://www.bentmachine.com/robot/elma/elbow/elbow.html

[davidhere40]

Thanks for the advice!
I should have mentioned, speed is very important :S
Normally, laborers can grab and move about one object every 2 seconds on average. The robot doesn't have to be capable of this. But the slower the robot is, the less cost effective it becomes. I would say the max I could probably allow for a grab cycle is 15 seconds. Much more preferable would be around 6 seconds.

Could a worm gear achieve this sort of speed? The path to the object won't be super simple or free of obstacles. So, it may require adjusting joints multiple times.

[Soeren]

Hi,

I should have mentioned, speed is very important :S
[...]
The path to the object won't be super simple or free of obstacles. So, it may require adjusting joints multiple times.

To sum up, you want to build a strong, obstacle avoiding, fast arm on a tight budget?
And I guess you want detectors in the arm to avoid those obstacles that probably will be in places that cannot be predicted?

Speed and price is inverse proportional and intelligent obstacle avoidance and price is inverse proportional as well, so there's a bit of a barrier here.
If you are trying to replace paid hands, it may be worth thinking in long term investment and throwing enough cash at it, to make it smart, strong and long lasting.

One way of getting the price down is to give it the environment needed, clearing obstacles rather than have it circumnavigate them - introducing a robot to a workplace usually means a bit of restructuring to give it the best possible conditions to make it worthwhile.

Are you sure a conventional arm (mimicking a human arm) is the best solution for what you need?
If you describe exactly what job you want the arm to do, perhaps someone can point you to a better/cheaper/simpler solution.

[davidhere40]

lol. Basically!
I think I found a solution though!
I'm developing a system for a previously unsolved problem. The application won't use sensors, but will use a camera with computer vision. The environment is unstructured and the obstacles are unavoidable. They can't be cleared because that's just the nature of the job the robot has to do. The space the robot works in is large... hence it needs a long arm. It's an industrial application that processes a lot of objects, so speed is important, but more arms can be used if necessary and cost effective. The industry also has rejected previously high priced robots because they can't do the same things that laborers do, but cost a fortune.
I'm afraid to mention the exact application, because I think its solvable with a little ingenuity and I don't want the whole internet to catch on just yet.

I can compromise on vertical movement. Vertical movements can be very slow, so I can use a worm-gear to move a jib-like piece up and down. The arm would be attached to the end of the jib. Here is an example of a jib I'm talking about.

Hague Camera Supports K8 Boom Jib

Then, in the X-Y plane (the horizontal plane) I will use rotary joints that move only horizontally. So, I'm thinking I could create a snake like arm with several joints stretching several feet out, where each joint can only move horizontally. The benefit of the rotary joints is that they only have to be strong enough at the joint to withstand gravity, but very little torque would be needed to move the joints. In between joints, I would use aluminum square bars. If light enough, it should be able to move easily and quickly. But, I can't be sure till I start testing ow it performs under its own weight and the weight of a 0.5kg object. Such a long arm may bend and twist because of the torque on the joints.

Any thoughts? Any ideas on how to construct light weight and strong rotary joints? I'm going to stop by a hardware store and see what ideas I can come up with.

[davidhere40]

I guess to make my latest idea clearer it would be closest to a SCARA robotic design. But, the length would be very long.
Ideally, I'd like it to be able to reach out about 10 feet using up to 10 joints.
Here is a SCARA robot example:
http://automation-thai.com/attachments/Image/scara%3D20long.jpg?template=generic&.jpg