Researchers from Baidu Research Robotics and Autonomous Driving Laboratory (RAL) and University of Maryland College Parker have launched an Autonomous Excavator System (AES) that can perform material loading tasks for a long time without any human intervention. Provides almost the same performance as an experienced human operator.
AES is one of the first unmanned excavation systems in the world to be deployed in real-world scenarios and work continuously for more than 24 hours, bringing industry-leading benefits in improving safety and productivity.
The researchers described their method in a research paper published in Science Robotics on June 30, 2021.
The corresponding author, Dr. Zhang Liangjun, head of Baidu Research Robotics and Autonomous Driving Laboratory, said: "This work proposes an efficient, robust and universal autonomous system architecture that enables excavators of all sizes to autonomously execute materials in the real world. Load tasks."
Excavators are essential for infrastructure construction, mining and rescue applications. In 2018, the global excavator market size was US$44.12 billion, and it is expected to grow to US$63.14 billion by 2026.
Given this projected market growth, construction companies around the world are facing a shortage of skilled heavy machinery operators, especially excavators. In addition, the continuation of the new crown pneumonia has exacerbated the crisis of labor shortage. Another contributing factor is that the dangerous and toxic working environment will affect the health and safety of human operators on site, including landslides, ground collapses or other excavation accidents, causing approximately 200 casualties each year in the United States alone.
Therefore, the industry is taking a scientific approach and seeking to create excavator robots to provide breakthrough solutions to meet these needs, making the development of systems such as AES a growing trend, and at the same time in manufacturing, warehouses And other robots implemented in autonomous vehicles.
Although most industrial robots are relatively small and operate in more predictable environments, excavator robots are required to operate under a wide range of hazardous environmental conditions. They must be able to recognize target materials, avoid obstacles, deal with uncontrollable environments, and continue to operate in difficult weather conditions.
AES uses precise and real-time algorithms for perception, planning and control, while adopting a new architecture to incorporate these capabilities into autonomous operations. Multiple sensors—including lidar, cameras, and proprioceptive sensors—are integrated into the perception module to perceive the three-dimensional environment and identify target materials, while using advanced algorithms such as dust removal neural networks to generate clean images.
Through this modular design, the AES architecture can be effectively utilized by excavators of various sizes-including compact excavators of 6.5 and 7.5 metric tons, standard excavators of 33.5 metric tons and large excavators of 49 metric tons-and is suitable for Various applications.
In order to evaluate the efficiency and robustness of AES, the researchers cooperated with a leading equipment manufacturing company to deploy the system in a waste disposal site. This is a toxic and harmful real-world scenario with a strong demand for automation. Despite the challenging task, AES can run continuously for more than 24 hours without any manual intervention. AES has also been tested in winter weather conditions. In this case, vaporization poses a threat to LiDAR's sensing performance. For a compact excavator, the amount of wet and dry material excavated is 67.1 cubic meters per hour, which is consistent with the performance of a traditional human operator. Dr. Zhang said: "AES has been stable and reliable for a long time, and the performance of human operators may be uncertain."
The researchers also set up ten different scenarios in a closed test field to observe the performance of the system in many real-world tasks. After testing various large, medium and compact excavators, AES finally proved to be equivalent to the average efficiency of a human operator in terms of the amount of material excavated per hour.
Dr. Dinesh Manocha, Distinguished University Professor in the Department of Computer Science and Electrical and Computer Engineering at the University of Maryland College Park, said: "This represents a critical step towards the deployment of robots with long-term operations, even in uncontrolled indoor and outdoor environments. ."
Looking to the future, Baidu Research Institute RAL will continue to improve the core modules of AES and further explore scenarios where extreme weather or environmental conditions may exist.
Baidu has been cooperating with several world-leading construction machinery companies to use AES to automate traditional heavy construction machinery. Dr. Haifeng Wang, Chief Technology Officer of Baidu, said: "Our goal is to use our powerful security platform to inject our powerful artificial intelligence and cloud computing capabilities to change the construction industry."
