Latest version of linear actuator robotic arm

[Hugh]

I have been (slowly) trying to design a robot arm with linear actuators for quite a while. The idea is very appealing to me because they have a great price/torque ratio, especially the holding torque with no current, but I have not really been able to find a configuration I'm completely happy with.
I think the arrangement of my latest iteration is about as good as I can make it, so I might look to make more incremental improvements from here.


The work against gravity is done by the linear actuators, the wrist is made from 3 Ax12 Dynamixels, the gripper is a mini linear servo and the base is a simple geared DC motor. The mechanical fabrication could use quite a bit of work, but it's a start.
I might replace some of the base gears with a belt drive to reduce backlash.

I show a bit more in this video:

[kshin]

I have been (slowly) trying to design a robot arm with linear actuators for quite a while. The idea is very appealing to me because they have a great price/torque ratio, especially the holding torque with no current, but I have not really been able to find a configuration I'm completely happy with.
I think the arrangement of my latest iteration is about as good as I can make it, so I might look to make more incremental improvements from here.


The work against gravity is done by the linear actuators, the wrist is made from 3 Ax12 Dynamixels, the gripper is a mini linear servo and the base is a simple geared DC motor. The mechanical fabrication could use quite a bit of work, but it's a start.
I might replace some of the base gears with a belt drive to reduce backlash.

I show a bit more in this video:

Great work Hugh! I really appreciate you sharing your success.
I noticed you have a controler with screen to to control your robot. It will be great if you can share with us how you are controlling the robot etc. Again, thanks for sharing!

Kevin

[Hugh]

The brains is a Raspberry Pi 3, connected to a 4.5" touch screen. There is an Arduino Nano behind the screen that reads a bunch of analog inputs (actuator positions, spring displacements) and the magnetic encoder and also uses PWM to control the motor driver for the base DC motor. This is connected to the Pi via USB and I have a simple custom protocol to send commands and transfers values.
Similarly, there is an OpenCM 904 on the wrist that reads the analog gripper position and spring displacement, and commands the Dynamixel bus and motor driver for the gripper. This too is connected via USB with a similar custom protocol to the Pi.


The main controlling program is an OpenGL program (written in "haxe" that compiles to c++ that is cross-compiled on a windows box to native RPi binary) that reads the USB serial points in separate threads. The touch events become mouse events that send commands to the individual micro controllers, or drives the actuators directly via the motor driver hat.
The Raspberry Pi camera feeds images into OpenGL textures that basically show up on the screen without the controlling program having to do too much.

Three of the USB ports are taken (2x serial, 1x touch screen) leaving 1 for a keyboard in development mode. Ultimately, I want to connect a neural-net accelerator to the last usb for image processing. The network port is at the top of the Pi, and this is quite convenient for programming and debugging.