A few days ago, Musk tweeted that the prototype of humanoid robot Tesla Bot Optimus (Optimus) will appear at the Tesla AI Day on September 30, and hinted that this Tesla AI Day will be shocking.
The release of Tesla Bot shows more that Tesla has become an AI company driven by "in-house chip + in-house algorithm".
01 "Optimus" will be released soon
At Tesla AI Day 2021, Musk unveiled the concept of Tesla Bot (Optimus), a humanoid robot. Recently, Musk has changed the Tesla AI Day, originally planned for August 19, to September 30, as the Optimus ("Optimus") prototype will be released at that time.
The "Optimus" (Optimus) will begin production as early as 2023 and will be used to fill the labor gap by performing jobs that are dangerous, repetitive or too boring for people to perform.
From the published parameters, Tesla Optimus humanoid robot main configuration:
Height 5 feet 8 inches (about 1.73 meters), head with a display to display information
Using Autopilot's camera as a visual perception sensor, a total of eight cameras
FSDComputer is used as the computing core
The neck, arms, hands, legs and torso are equipped with a total of 40 electromechanical actuators
Equipped with multi-camera neural network, neural network-based regularization, automatic tagging, algorithm training, etc.
Built with lightweight materials and a top speed of 5 mph
The robot will allegedly reuse Tesla's technology in the field of autonomous driving, and is expected to use Tesla FSD's computing modules and solutions. According to Musk, to solve the problem of autonomous driving, real-world artificial intelligence needs to be addressed first, so from a product perspective, it is not surprising that it uses Tesla's most advanced AI technology in its humanoid robot.
Tesla has developed its own super meter for autonomous driving - Dojo - and its own AI training chip, the D1, which is an important part of the Dojo architecture. As each D1 chip is seamlessly connected together, the latency between adjacent chips is extremely low, and the training module maximizes bandwidth reservation, and with Tesla's own high-bandwidth, low-latency connector, the arithmetic power is up to 9 PFLOPs (9,000 trillion times).
Tesla's FSD requires the assistance of artificial intelligence algorithms both in perception and in cognitive decision making. In terms of perception, the FSD is equipped with the neural network-based Telsa Vision vision processing tool to achieve deconstructive analysis of the driving environment. In terms of analysis and decision making, Tesla has also introduced reinforcement learning algorithms to leverage performance improvements.
With the expected release of prototypes of Tesla's humanoid robot, Tesla is expected to upgrade to two core product lines in the future: the Model series of electric cars and the Optimus series of robots, with Musk stating that "the value of Optimus will surpass Tesla's electric car and fully automated vehicle business".
It is foreseeable that the high adaptability of humanoid robots can break through the restrictions of industrial, commercial and household scenarios, and integrate the market space of various applications, which is expected to become a new generation of intelligent terminals after PCs, cell phones and intelligent electric vehicles.
02 Human-robot interaction enters a new era
The development history of humanoid robots can be traced back to the Da Vinci era at the end of the 15th century, when an old "humanoid robot design sketch" drawn by Da Vinci - a "machine" driven by wind and water power - inspired the development of human-robot interaction. This inspired the human fantasy of developing "humanoid robots".
In 1927, Westinghouse manufactured the world's first humanoid robot, "Televox," which could not walk, but could lift the receiver to answer a phone call and control a simple process by operating a switch based on the signals received, in addition to answering a few questions.
In 1937, Westinghouse built the "Elektro the Motorcycle Man", considered the first true humanoid robot. 210 cm tall and weighing over 120 kg, Elektro was able to perform 26 different daily activities, including waking up, talking, counting and smoking.
In 1961, the first digital and programmable robot, Unimate, was installed on the General Electric assembly line and became the basis for many of the industrial robots still in use on assembly lines today.
In 1972, Waseda University developed the WABOT-1 robot, the world's first full-size humanoid robot. It was able to communicate in Japanese, navigate rooms and grasp and transport objects. The later WABOT-2 was able to read music scores and play an electronic keyboard.
Since 2000, the direction of humanoid robot development has gradually matured. In terms of technology and product maturity, Boston Dynamics is the industry leader.
In 2005, Boston Dynamics researched and designed the quadruped robot BigDog for the U.S. Army, and since then, it has developed HANDLE for logistics scenarios, and the robot dog SpotMINI. 2019 saw the latest version of its humanoid robot Atlas, which has considerable maturity and is capable of completing more complex tasks in limited spaces.
03 Core components are the key to robotics
In all robot development, the challenge of humanoid robots is recognized as the most difficult in the industry.
From the current scope of applications announced by Tesla, the humanoid robot is a technological upgrade of the service robot, involving the integration of various technologies such as autonomous driving, visual navigation and sensor technology, and the mechanical industry chain mainly involves servo motors, reducers, control systems, drives and machine vision.
The robot industry chain is mainly divided into three aspects: upstream core parts R&D and production, midstream body manufacturing and system integration, and downstream specific applications in various industries.
Among them, core components such as reducers, servo motors and controllers are the core barriers of the whole robot industry, and are also the parts of the industry chain with the highest cost and added value.
The reducer is mainly used for each joint of the robot, and is responsible for converting the high-speed running power output from the servo motor into low-speed and high-torque motion. Therefore, if humanoid robots want to realize the freedom of movement close to human body, the amount of reducer will increase by orders of magnitude.
According to the different transmission principles, robot-specific reducers can be divided into two kinds of RV reducers and harmonic reducers. At present, China's precision reducers rely heavily on imports, and domestic manufacturers are still in the stage of continuous maturation and improvement of production technology. According to the data of MIR DATABANK, about 75% of the global precision reducer market is currently occupied by Japan's Nabteske and Hammonaco. Among them, Nabteske produces RV reducers, accounting for about 60% of the share; Hamer Nacor produces harmonic reducers, accounting for about 15% of the share.
Servo system is the robot's power system, responsible for driving the robot's movement, and consists of servo driver and servo motor. Among them, the servo driver is responsible for the position, speed, torque and other controls; servo motor is the engine that controls the operation of mechanical components in the servo system, converting the received voltage signal into torque and speed to drive the control object.
As the heart of the robot, the controller is responsible for planning the robot's movement, issuing and transmitting action instructions through hardware and software, controlling the robot's movement position, attitude, trajectory and operation sequence, etc. to ensure the normal operation of the robot system and make the robot achieve the required technical specifications. At present, ABB, KUKA and other major robot manufacturers' controllers are independently developed on the basis of a common multi-axis motion controller platform.
Robots are often regarded by countries as a measure of national innovation level and industrial capacity, and humanoid robots are regarded as the crown jewel of robots. The launch of Tesla's humanoid robot is expected to accelerate the development of the global service robot industry.
04 Market space can reach millions of levels
Tesla Optimus price is expected to step into mass production level, and humanoid robots open the global trillion blue ocean market. According to Elon Musk's TED talk in April 2022, the unit price of Optimus will be "not horribly high, perhaps lower than the price of a car". price range for reference.
According to a report released by McKinsey Global Institute, it is expected that by 2030, 15% of the global labor rate will be replaced by robots. Among them, India, China, the United States, Germany replacement ratio of 9%, 16%, 23%, 24% respectively. However, not all robots need to be made in human form. According to the application scenario, some robots in service robots and special robots are suitable to be manufactured into human form.
Currently, robots are shouldering more and more jobs and assuming more and more roles in industrial production and daily life, and their importance is self-evident. The robot industry will continue to develop rapidly in the future, but there are still many pain points to be solved in order to reach large-scale commercialization.