Why are Commercial Robotic Arms so Expensive?

[Flaviu_I]

I tried to see why the Willow Garage PR2 robot is so expensive and here is what I found in their price comparison explanation:

"$160,000
Barret Technologies WAM Arm is a 7DOF back drivable current-controlled robotic arm with integrated drive electronics comparable to the PR2 robot arm."

"Gripper: $35,000
Barret Technologies Gripper is a gripper capable of grasping objects comparable to the PR2 gripper. The Barret Gripper has more degrees of freedom, but lacks the accelerometer feedback and 44-element pressure sensors of the PR2 gripper. Pressure sensors for the Barret Gripper significantly raise the price of the Barret Gripper, but that is not accounted for in this comparison."

All I can say is WTF? A decent car costs less than a gripper? A well equipped Porsche costs less than a robotic arm?

I understand that there can be explanations for the economics of price setting (e.g. Barret Technologies does not price according to marginal cost, since there are few entities that buy such robotic products, so they need to recover their upfront investment from fewer customers, so they price the unit higher)

What I want to know is how hard would it be to replicate --at slightly lower performance levels-- the technology in those grippers and arms. Would it be that hard to build a high quality and fidelity robotic arm for, say $10000? A quality gripper for $5000?

What are the main technological impediments to building such high quality products?

[newInRobotics]

It's not the price of manufacturing robotic arm that makes it expensive, it's the price of development. Making prototypes cost the most, as whole factory has to be equipped to make 1 prototype, if something is wrong and design has to be altered, so as tools used to make the robot. Developement requires thousands of man hours as well - imagine small team of 20 engineers working on development and getting £25 an hour; do You see my point? Another factor, as You mentioned Youself, is that robotic arms are used by very small group of people, hence manufacturer/developer has to keep prices up in order to earn something.

Regarding replicating for lower price - not likely. If You decide to make one Youself (DIY style) - precision will be horrible. If You decide to design robot (that uses current technology) Yourself and hand over design to someone else to manufacture, price will be horrible (it will exceed price of ready made robot many times, as in essence Your robot will be prototype with custom parts, plus You need only one robot, making You unacceptable customer unless You pay very well).

[Flaviu_I]

If You decide to make one Youself (DIY style) - precision will be horrible.

That is the kind of detail I am interested in : what is the difficulty in making precise robots? Are there special actuators that are required? If so, are such precise actuators available commercially?

I am also interested in what is the difficulty of using somewhat precise sensors in a robotic gripper. Are there no commercial and relatively inexpensive (compared to the insane $35000 price tag of the gripper I mentioned) options for sensors?

I understand that designing a precise robot is very difficult and expensive -- but I want to know if there are other, more "physical" constraints on building relatively inexpensive (say, with a price comparable to a car) robots.

[newInRobotics]

That is the kind of detail I am interested in : what is the difficulty in making precise robots? Are there special actuators that are required? If so, are such precise actuators available commercially?

I cannot give You any specific numbers in this case as I never wanted to build a clone of an industrial robot, but from what I gather, developing high precision robots involves not only using actuators that fall into required tolerance range, but also developing alloys that are light and durable, shaping those alloys so that minimum amount of material is required to, again, achieve lightness and durability (it's not enough to have 2 pieces of wood connected with servo in the middle to get accurate movements). Then there are control algorithms responsible for error detection and correction. There are sensors all over the robot to provide feedback on various events.

Regarding Your question about actuators and their prices - no, no special actuators are required, they are plain stepper motors in most cases. Think of this - the bigger the motor normally the lower its precision is, in other words, if you have robot arm which length is 10cm and small stepper (to move the arm) has tolerance (error) of +/-0.1° - then the end of the arm has an error of +/-0.174mm; now, imagine big robot arm of 2m with big stepper motor of same +/-0.1° error - then the end of this arm has error of +/-3.49mm. Most industrial robotic arms have at least 3 actuators, so if every motor has movement error of +/-0.1° - big inaccuracy occurs. There is need to have big motor which has less rotational error than the small motor, here comes the price. You can get actuators used in industrial robots (maybe not exact ones, but of similar specs.), however they will be more expensive as You need only 3 or 4 of them, where companies like ABB buys a lot more on one go, hence one unit to them costs way less than to You.

I am also interested in what is the difficulty of using somewhat precise sensors in a robotic gripper. Are there no commercial and relatively inexpensive (compared to the insane $35000 price tag of the gripper I mentioned) options for sensors?

$35000 is not the price of the sensor - it's the price of whole product - same rules apply as previously stated. Sensors are fairly cheap these days, however, the ones used in grippers are not from Maplins or Digi-Key, they are most likely developed only for the company that uses them and fit with their designs. Another factor is precision; if You want Your robot capable of crushing an elephant to handle something fairly fragile (like windshield of the car), or for precision welding operations, You most likely want it to be able to detect smallest pressure, or angular rotation, or vibration, etc, again - that's were the price comes from.

I understand that designing a precise robot is very difficult and expensive -- but I want to know if there are other, more "physical" constraints on building relatively inexpensive (say, with a price comparable to a car) robots.

It all depends how much precision You need, how much time You have, how many good friends, who work with various manufacturing equipment, You have and how willing they are to help You, and for how much they are willing to help You. If plan is to use hand tools to produce robot made of wood or aluminum, or carbon fiber using few high power steppers, then yes, it is doable, but will take loads of time, main price factor will be precision of components, and Your robot will be as precise as Your power drill is and also as much as You chose to invest into components. End result wont be anywhere near what You can see in the video below:

http://www.youtube.com/watch?v=SOESSCXGhFo

[Soeren]

Hi,

[...] if You want Your robot capable of crushing an elephant [...]

I think we have to report you to PETA (People Eating Tasty Animals)

[Flaviu_I]

Thank you for the answers.

So it is likely that the cost of the design is one of the biggest upfront costs -- the rest are important costs, but not significantly different from the costs of a car.

A precise robot probably needs lots of control theory inspired feedback loops between sensors and actuators, which is probably what gives it part of the precision. Probably just precise actuators by themselves would not be enough.

In the end, it is just the lack of mass production that makes those robots so expensive -- I'm raising a glass to a future with a robot in every home!

[davidhere40]

Yep, I think your last reason is the biggest one. Economies of scale. If you're not going to make hundreds of thousands of them, you won't spend time automating and optimizing production. Your costs will also all be supported by a small number of sales, meaning that each sale must be larger to cover costs. Whereas, if you sell a million, you can divide all the one time costs up among that million and make each more efficiently.