A Zhihu user raised a question:
Why don't robots use ball screws to drive rotational degrees of freedom to reduce the cost of speed reducers?
Problem Description
Multi-stage reducers for mechanical servos are inefficient, and cycloidal pinwheels or harmonic reducers have high material requirements and are expensive. Why not use linear actuators such as ball screws, which are highly efficient and extremely low cost, like biological joints, to convert linear motion into rotational motion of the robot's joint axis?
Here's a diagram of a common harmonic reducer. The thin-walled cylindrical structure in the middle is the flexspline, and the harmonic generator in front isn't actually a perfect circle; it deforms once with each rotation. The force from the meshing of the flexspline's teeth with the rigid gear is transmitted to the outer circular spline, causing one tooth to rotate, achieving a very high reduction ratio and lightweight, efficient transmission. It's clear that this structure places high demands on the flexspline material, making finished products generally very expensive.
A typical KUKA six-DOF robot has the first three degrees of freedom on the base and arm, which have a large load-bearing capacity and high load requirements. The latter three degrees of freedom are on the wrist, which is mainly used to achieve omnidirectional adjustment of the end effector.
So, to summarize:
Pros :
Typical ball screws have an efficiency of over 90%, which is not worse than high-efficiency, high-reduction-ratio reducers such as harmonic reducers.
The ball screw, with multiple balls bearing force in parallel, seems much more reasonable in terms of impact resistance than the rotary reducer, which relies on high-quality materials and a small number of tooth tips to withstand impact, thus resulting in high costs.
Aside from the advantages of traditional rotary reducers, such as their compact structure and convenient assembly design, ball screws are simple, practical, widely applicable, and very inexpensive. They hold promise for use in the main arm joints of robots, significantly reducing the cost of multi-degree-of-freedom robots.
Following this line of thought, imagine that the resulting robotic arm would look like this, full of servo electric actuators (hydraulic experts, please consider this question as "Why not use servo electric actuators to replace hydraulic cylinders for miniaturized equipment?").
Cons :
Some speculations :
One possible answer is that in the past, ball screws had singular angle problems in the rotational motion of the shaft through lever arm conversion, and the nonlinearity of the control process made controller design difficult. However, firstly, the mathematical model of this process is simple, and secondly, the development of various advanced control methods, especially active closed-loop assistance such as vision, may help solve this problem. Perhaps, due to the limitations of the existing robotic arm industrial system, such a new architecture is difficult to use.
The second possible answer is that the motion transmission of ball screws is irreversible, which poses difficulties for applications in some fields (for example, a power outage could render the entire structure like a zombie, stuck in mid-air and unable to recover, which could be fatal in some situations). However, methods such as monitoring structural strain and actively compensating for it can somewhat mitigate this difficulty.
I suspect another possible answer is that this application requires ball screws to withstand bending moments, which is obviously unreasonable. Is it necessary to improve this?
Another possibility is that the almost irreversible motion transmission of the lead screw kinematic pair creates difficulties in teaching and reproducing the motion. Of course, there are solutions, but they differ greatly from typical industrial robot usage habits.
In addition, the error caused by multiple links is also relatively large. In Delta robots, the error is canceled out by parallel connection, but the error increases for lead screw and link driven mechanisms.
WorkEnvelope is a major flaw.
Zhihu user Wang Zhe offered his insights on this issue:
Wang Zhe clarifies
Why not use a linear-to-rotary mechanism? The primary issue is the working range! In fact, the answer is already quite clear from the questioner's description.
This is a schematic diagram of an excavator, an example of a linear-rotating mechanism provided by the questioner. The working area is roughly marked as shown in the following figure:
A typical industrial robot workspace looks like this:
The ratio of the working domain to its own volume is much greater than 1 for robots, while that for linear-rotary mechanisms is much less than 1. Moreover, multi-joint robots can reach a very wide range in the spatial domain and are extremely flexible.
The drawbacks of linear-rotary mechanisms are fatal. Industrial robots, in many cases, are actually inferior to dedicated hard automation in terms of cost and efficiency. However, their popularity lies in their flexibility and versatility, requiring less interference verification and mechanical and structural design during the design phase.
Flexibility and versatility are built upon a large work domain and extremely free accessibility, for example:
When performing welding processes, you need to deal with many workpieces of irregular shapes. If the robotic arm is severely limited in its reach, no one will adopt such equipment.
For machinetending applications, a simple schematic diagram is sufficient, as everyone knows that once the robot is placed, interference will not be a major issue; only minor adjustments to the posture and installation position are required.
Try using an excavator-like mechanism in such a scenario; it would be an absolute nightmare during the design phase.
In fact, just one of the aforementioned drawbacks would be enough to shut down that kind of institution. Of course, there are other problems as well:
Secondly, there is space.
A linear-to-rotary mechanism is essentially a lever, and a lever requires additional space on the arm. Leaving aside the issue of the working domain, this increase in space will lead to a large amount of motion interference.
Thirdly, there are problems caused by lubrication.
Ball screws require lubrication, and because they frequently extend and retract, some mechanisms are exposed and carry grease. Contamination is one issue, and regular maintenance is another. Modern robots only require maintenance every one to two years because of the sealed design of the reducer, preventing excessive grease dissipation and contamination.
Fourth is weight.
Don't underestimate this little bit of weight. For multi-stage transmission in robots, the inertia and damping caused by the weight will be magnified many times over.
Of course, there are other issues, such as control algorithms, but these are minor compared to the aforementioned shortcomings, especially the working domain problem.
Clarification and Supplement
In addition, I'd like to answer the questioner's supplementary question.
Every mechanical structure, everything has its advantages and disadvantages, but for any specific application, these advantages and disadvantages have varying degrees of importance, that is, they have different weights. Specifically, in the application of industrial robots, the disadvantages of this structure are fatal and decisive.
Moreover, this defect is at the principle level.
This is a schematic diagram of the robot's axes (each major robot manufacturer uses different names for each axis; SLURBT is Yaskawa's name, FANUC calls them J1-J6, and ABB and KUKA generally call them axes 1-6. For convenience, we will use Yaskawa's STU system here).
First, we will not discuss the three axes S, R, and T in the diagram, but only consider the three axes L, U, and B.
Look at my sketch:
Each telescopic rod—the transmission mechanism for rotation—in principle, operates within an angle range α of 0.
In practice, the lead screw thrust F is decomposed into a radial force f2 along the rod and a normal force f1 perpendicular to the rod due to the angle b between the lead screw and the drive rod. f1 is effective, while f2 is ineffective. f1 = F * sinb. Since the angle b is at its extreme positions near 0 and 180 degrees, sinb will approach 0 infinitely.
Due to other mechanical limitations, angle a can reach 30.
The angle range of a typical industrial robot is shown in the figure below.
With three axes, each axis has its range of motion halved, resulting in an extremely compressed reach domain. As I mentioned, the work domain and flexible reach capability are the fundamental values of articulated robots.
Your design is equivalent to saying that you have a design that can reduce the cost of a ship by half, but at the cost that the ship can no longer float on water. There are no "advantages and disadvantages" to talk about here.
