Foreword: In recent years, with the rise of service robots, robotic arms are also entering our daily lives. Service robots will continue to be subdivided in application scenarios and service methods, going deep into various fields, and have broader application prospects than industrial robots. , the industry scale may become larger. And in recent years, some low-cost, small, "desktop version" robotic arm products have appeared. I believe this will be an important direction for service robots!

Speaking of robotic arms, you may think of this:

Yesterday, the vast majority of robotic arm applications were in factories, such as spot welding and painting in automotive plants, packaging and filling in chemical and pharmaceutical plants, and work in high-risk environments such as nuclear power plants. These robotic arms are thick and powerful, but a bit bulky. Most of them are kept in cages to perform simple repetitive tasks.

Unlike standardized industrial robots, which only undertake production and manufacturing tasks, service robots have many downstream applications and have only one role - to serve all mankind, in addition to production and manufacturing work.

After a brief thought, you will find that the world's robot industry is like an ink painting. Industrial robots stand tall and dominate the global manufacturing industry, while service robots are pouring into everyone in life in an inclusive and broad form. A tributary.

In recent years, with the rise of service robots, everyone hopes that robotic arms can also enter our daily lives. So there are low-cost, small, so-called "desktop" robotic arm products.

Today’s robotic arm looks like this:

As for what it is used for? Pick up a bank card, write a love letter, cut vegetables... (Editor will quickly add!) Then you need to use your own imagination. Many friends say, "The accuracy is poor, the speed is slow, the power is small, and the eggs are useless!" "Poor accuracy, slow speed, and low power" is true compared to hundreds of thousands or even millions of industrial robots, but I think the desktop version is not used for physical work after all, it is used to open up imagination, then You need to design algorithms and practice programming! What? Won't? Walk slowly without sending. . . . . . What? Need a delivery? Then continue with the little popular science below.

The robotic arm is composed of many connecting rods (metal rods) and joints (electric pivots). A typical robotic arm has a base, an end effector (such as a clamp, suction cup, or pen) and six links and joints between them, so that the robotic arm can achieve various tasks within its range of motion. Arbitrary three-dimensional position and orientation (pose). At present, the control of the zekeep robot still directly controls the motors of each joint. This is relatively simple, but in fact, what you really need is to move the end gripper to the desired position, such as picking up an object in the distance.

But here comes the problem. Motors 1-5 are controlled manually one by one, which is extremely cumbersome! Now there is a solution. In the first step, the operator grasps the end of the robotic arm, places it at the desired position, and presses the operating button on the console to record the position and posture of the robotic arm at this time. A series of positions are recorded in sequence. The pose is then used as a reference point for robot programming, such as repetitive movements in pick-and-place tasks. The robot arm will remember the angle and position of the joints, program it automatically, and then repeat it. This is a teaching demonstration of the small robot arm, which can easily allow novices to get started using the small robot arm faster, saving a lot of work. The time cost of learning.

Regarding the future of small robotic arms, in order for us to make better use of robotic arms, we need to learn and understand the uses of high technology in depth.

1. Improve motor technology

The motor is currently a bottleneck of the robotic arm. It is not only small in power, but also high in cost, which significantly affects the final cost of the robotic arm.

Motors are related to the very different working mechanisms of joints and muscles in biological systems. Our human arm joints are both light and extremely flexible. The flexibility of each joint and the hardness of the muscles can be controlled, so in fencing, you can keep your muscles relaxed while parrying, and easily move the foil in your hand to the side; or in an arm wrestling contest Toughen the muscles in your arms to keep them strong and not knocked down. Our joints also move very quickly. By comparison, electric motors are heavy, energy-intensive, and stiff. If you add a gearbox to the motor, you can increase the output torque, but it also reduces the rotation rate and makes the output shaft difficult to rotate - you can't twist the arm by hand without exerting some effort, and you can't turn it by hand. The output shaft of the gearbox. If you build a small robotic arm with lightweight motors inside it, it's completely impossible for the arm to lift a coffee cup, even if the arm is as light as a human arm. And this task is too simple for robots that are required to help the elderly out of bed at home and assist them to go to the bathroom. To complete these actions, the robot must have flexibility, elasticity and considerable strength at the same time.

For a while, many people believed that the future of robot motors should be systems that work like human muscles, with nickel-titanium alloy metal muscle fiber technology being hailed as the next revolutionary advancement. This kind of metal fiber can stretch and contract with the change of electric current. People have designed ingenious devices to make multiple clusters of similar fibers work in parallel to carry weight, and to extend its range of movement from a few millimeters to several centimeters and longer distances. However, challenges with overall energy efficiency and fatigue have limited the application of such metallic muscle fibers to certain areas, such as medical robotics. This is an area where a tightly controlled environment makes working ideal.

Nonetheless, recent work on novel motor designs has resulted in highly dynamic, controllable motors. Researchers are developing a joint that integrates motors and adjustable springs, a system that can delicately respond to external impacts. More sophisticated motors already have built-in pressure sensing, and they can achieve extremely fast control. The motor's electronics can sense the external pressure acting on the motor and react to these pressures in real time to output any desired stiffness and tenacity. One day, this will allow robotic arms to perform strong and safe handshakes, origami, and even crack eggs while keeping the yolk intact. One of the reasons why motors have advanced is that although they do not follow Moore's Law themselves, they benefit from the continuous advancement of embedded control circuits and software, so that Moore's Law can influence future motor technology in another way.

2. Learning-based robotic arm control

Current robots also follow the sense-plan-act paradigm. The robot observes the world around it, constructs an internal model, formulates an action plan, and then executes this plan. This is OK in a regular environment, such as visual servoing in the traditional sense mentioned above. But in various messy natural environments in the real world, robotic arms are really messy. So we still need to train robots to reliably handle complex real-world situations!

3. Increase the visual sensor control speed

Humans and animals typically do only very little advance planning in reality, relying mainly on highly developed intelligent feedback mechanisms that use sensory information to correct errors and compensate for perturbations. In tennis, for example, the player constantly looks at the ball and racket, adjusts his hand movements, and ultimately catches the ball in the air. This feedback is fast, efficient and, crucially, can correct errors or unexpected perturbations. The robot uses the image obtained by the visual sensor as feedback information to construct a closed-loop feedback of the robot's position, also known as visual servoing. In recent years, visual servo control systems have been a hot topic in the field of robotics research.

Based on the different ways of utilizing information, visual servo systems can be divided into two categories: one is position-based visual servo system (PBVS), and the other is image-based visual servo system (Image-Based Visual Servo System). Servo system, IBVS). According to the different camera positions, visual servoing can be divided into two types: global vision and local vision (eye-in-hand).

The advantage of global vision is a wider field of view, but the disadvantage is that it has low resolution and target occlusion problems; while the characteristics of local vision are exactly the opposite of global vision. Therefore, some people have proposed a visual servo control method that combines global vision and local vision. The performance of visual servo control is closely related to the calibration errors of cameras and robots, and the accurate calibration of the model is a complex task. Therefore, the calibration-free model of visual servo control has attracted widespread attention. Its main advantage is that it does not need to know the robot model and calibration camera parameters, has adaptive capabilities, and can estimate the Jacobian matrix online.