I am thinking about building a small robotic arm with 4 small servo motors and an arduino uno to apply basic neural networks concepts.

Is it a good idea to use a hand made robotic arm to learn more the power of neural networks?

Thank your for your time and Merry Christmas

  • 1
    $\begingroup$ I voted to close this question because life questions (what robotics project to construct) are off topic. That said, from a practicality point of view, I would go with a commercially available, off-the-shelf arm to ensure that you're not introducing errors with a home brew arm that would interfere with your algorithm. $\endgroup$
    – Chuck
    Dec 26 '15 at 0:53
  • 1
    $\begingroup$ I agree with @Chuck that the question is off topic - However, to build a robot arm from the ground up, you could have a look on eBay and purchase these brackets, along with some cheap MG996R servo clones, which works out cheaper than the Robot arm kits from which they are made. $\endgroup$ Dec 26 '15 at 4:13
  • 1
    $\begingroup$ Have a look at my blog - I've looked into the various kits available. I can't help with the neural networks though :-) $\endgroup$ Dec 26 '15 at 4:13
  • $\begingroup$ I am not sure a roboric arm is the best way to learn about neural networks. I have been doing robotic manipululation for many years and have never heard of such a thing. That being said, why don't you describe your project in more detail so we can help better. $\endgroup$
    – Ben
    Dec 26 '15 at 18:33
  • $\begingroup$ Basically i want to build a robotic arm to implement later a neural network to try to simulate an automatic pid tunner for example $\endgroup$ Dec 27 '15 at 17:15

I Would vote no.

Building a robotic arm yourself is quite difficult, but like others said, is possible. However, this will mean that you spend a lot of time building this and not focussing on your goal of learning about neural networks.

If you want to start with neural networks, I would vote to start with image recognition, or basic numerical problem solving, all of which can be done in software, without the need for mechanics.

If you want to test a neural network on a robot, I would vote to take an existing robot platform (can be an arm, or a rolling platform, or anything else) and load your neural network on the controller.

I think this way, you will learn much more in less time. But if you want to learn how to build an arm, by all means, go ahead!


Building an arm yourself could be a bit tedious. If your in a mechanical background then yes it might be of some help but if your in embedded or electronics then its better to order a already built frame on which you can add the servo in place. Some have the servos in them itself.

I worked in a project where I made a robotic arm controlled by user given angles. I used '16-Channel 12-bit PWM/Servo Driver-I2C interface-PCA9685 for Arduino' to connect my microcontroller to the Robot. And the Controller is used to set the match register value present in the interface boards timer unit. This in turn produces a PWM which is given to the servo. The timer match register value changes depending on the angle you need the arm to move and thus the PWM also changes according to the match register value. So this should be a very cheap and basic robotic arm project I guess!


Not the answer you're looking for? Browse other questions tagged or ask your own question.