View Full Version : Experiment with linear actuators

09-29-2014, 12:59 AM
I was looking to build a robot arm with a reasonable lifting capacity (couple of kilos), and had difficulty finding something stronger than robot servos but cheaper than a full-spec industrial arm. The solution I have come up with is to use linear actuators for the main lifting components, and I think I will probably use Dynamixels for the wrist/gripper.
Here is a video of in action (sorry about the lighting):

I think this approach is pretty promising.

09-29-2014, 06:52 PM
Linear actuators are great when you don't have to worry about humans. They are strong, precise, and fairly inexpensive (to buy or build), but slow and non-backdrivable (definitely electric lead screw versions; pneumatic and hydraulic can be a bit more forgiving under some conditions). This means that if anything gets in their way, then they will keep going until they destroy their internal drive system, the structure to which they are attached, and/or anything in their way. Of course, the very expensive Dynamixel Pro servos use a non-backdrivable cycloidal drive, so should be fine as long as you include plenty of protection against contact with foreign objects (padding and/or FSR skin for contact sensing (http://www.hsi.gatech.edu/hrl/project_open_source_whole_arm_tactile_sensing.shtm l)).

09-29-2014, 11:28 PM
Good points - the laws of robotics suggests that I should not try to harm anyone. I saw somewhere that someone had added a stiff spring with displacement measurement to a linear actuator to create compliance and force measurement for both safety and control.
This also seems like an interesting idea.

09-30-2014, 03:38 AM
You could also use springs and cables to transmit the forces with compliance while keeping the linear actuators safely tucked inside a housing. The DLR hand uses dyneema fibers, motorized capstans (position control), and motorized roller-arms (real-time adjustable cable tensioners and torque sensors) to make a very fancy and durable arm/hand - they can hit the fingers with a hammer, then put the hammer in the hand to drive nails. The 'low-cost compliant 7-dof robotic manipulator' from a group at Stanford (circa 2011) uses stepper motors for the big timing-belt/cable-driven joints in the shoulder and elbow then uses dynamixels for the hand and wrist.

I bought two spools (500m black and 300m blue) of generic 100lbf dyneema multi-strand fishing line for ~$25 to experiment with a cable-driven arm but still have not gotten around to using it for robots. Have to be careful with the stuff; it is so thin and strong that it can rather easily cut through flesh and itself when tensioned.

02-06-2015, 10:39 PM
Here's a little update to the design using the position feedback to allow control via USB.