Several Chinese departments including the National Development and Reform Commission recently issued a work plan for improving the statistical and accounting system for carbon emission, highlighting the establishment of a sound carbon footprint management system.
As an important concept in evaluating low carbon, carbon footprint is used to identify and evaluate the greenhouse gas emissions of products.
The work plan mentioned that national standards such as the general requirements for product carbon footprint quantification should be formulated and issued, and the application of green electricity certificates should be strengthened in the carbon footprint accounting system of key products.
The work plan also includes carrying out pilot projects for labeling and certification of product carbon footprint, and the establishment of internationally influential accounting, evaluation and certification agencies for product carbon footprint.
Relevant departments proposed goals that by 2025, the national and provincial carbon emissions annual reporting and bulletin systems will be established, a set of accounting standards of carbon emissions and product carbon footprint for industries and enterprises will be released and implemented, and the construction of a product carbon footprint management system will make positive progress.
China has announced that it will peak carbon dioxide emissions before 2030 and achieve carbon neutrality before 2060.
China to enhance statistical accounting system for carbon emission
China to enhance statistical accounting system for carbon emission
Advances in robotics algorithms and the accumulation of data have enabled China's humanoid robots to rapidly evolve, progressing from being able to quickly recover from falls to playing a crucial role in various practical applications.
A recent algorithm developed by the Shanghai Artificial Intelligence Laboratory, in collaboration with institutions including Shanghai Jiao Tong University, allows humanoid robots to rise quickly and stably from various positions.
"We are likely the first in the world to have developed an algorithm capable of standing up on generalizable terrain and arbitrary terrain. Previously, robots could only rise from fixed terrains in fixed postures with fixed procedure. Our technology, which advances through interaction, now allows them to stand up across various scenarios and terrains with various postures," said Peng Jiangmiao, a scientist at the Shanghai Artificial Intelligence Laboratory.
Imitating human movement is a vital path for robots to acquire skills. The laboratory has developed a new remote operation system for data collection, which includes exoskeletons for the upper limbs and pedals for the lower limbs, enabling robots to precisely replicate the operator's movements.
"This system allows a single operator to control both the upper and lower limbs of the humanoid robot, improving the efficiency of data collection. In the future, it is hoped that this cockpit could be used in hazardous environments, such as fire or toxic gas areas, where operators can remotely control robots," said Ben Qingwei, a joint PhD student at the lab.
Data is crucial for training large models. AgiBot, an innovative company dedicated to developing AI-powered robots, has a 2,000-square-meter data collection center, where 100 humanoid robots are currently undergoing a range of training exercises. The center has made over one million pieces of training data publicly available.
"We've designed multiple large categories of scenarios, including home, retail and service, and industrial settings. The data collected from these varied scenarios will be gathered to train our robots' large models," said Yao Maoqing, executive president of Agibot's research development. Unlike typical language models, which can rely on the vast amounts of text data already available online, humanoid robots face a shortage of real-world data. For example, performing a simple task like pouring milk requires nearly 100 high-quality data points.
"At the end of December last year, we released a dataset of one million robot trajectories to the public, hoping to advance progress in embodied intelligence in both academic and industrial fields. Going forward, we will continue to optimize costs, including the overall cost of robots, and improve the efficiency of data collection," Yao added.
China's humanoid robots evolve with technological, data breakthroughs