Universities and Researchers Team Up to Bring Fully Autonomous Machinery to Construction Sectors
Baidu Research Robotics and Auto-Driving Lab (RAL) joined University of Maryland to create the first "fully autonomous" excavator system.
Robotics Research Companies and Universities are Teaming Up
This new autonomous excavator system is capable of digging at rates similar to that of a human operator. The excavator system is still in the testing phase but has shown promise by completing a 24-hour digging task without human intervention.
Video used courtesy of Baidu Research
Baidu Research Robotics and Auto-Driving Lab bring computer vision and innovative technology to autonomous driving, industry, and service robots. RAL strives to bring lab research results to the global market through expertise in:
- Computer vision
- Machine learning
- Simulation and systems
- Studying holistic robotics solutions
Researchers at the University of Maryland, Baidu Research Robotics, and Auto-Driving Lab published a paper on June 30, 2021, in Science Robotics to document the results of their efforts and their research process.
New Autonomous System Uses LiDAR, Control Algorithms, and Sensors
The autonomous excavator system (AES) combines different technology to create a system that can keep up with the demanding environments that excavators often operate in. It uses perception, planning, and control algorithms in real-time to adapt to a continuously changing environment. LiDAR, cameras, and proprioceptive sensors are all part of a system of sensors used to provide feedback to the excavator during operation.
An overview of the hardware in the AES system. Screenshot used courtesy of Baidu Research
"This work presents an efficient, robust, and general autonomous system architecture that enables excavators of various sizes to perform material loading tasks in the real world autonomously," said Dr. Liangjun Zhang, corresponding author and the Head of Baidu Research Robotics and Auto-Driving Lab.
The sensors create a perception module capable of understanding a 3D environment and can be used to identify different target materials. An advanced algorithm is used for creating a dedusting neural network to generate clean images. The system is designed to be modular, capable of being used on everything from compact excavators to 49 metric ton large excavators.
Testing and Use
This AES was tested in a waste disposal site, which constitutes an environment traditionally harmful to equipment operators. Researchers went to a leading manufacturer of excavators to better understand the usefulness of the system. They tested the system on excavators of different designs and sizes in challenging environments. The test provided helpful results, and the AES could operate without human intervention for 24 hours.
The autonomous excavator system (AES). Image used courtesy of Baidu Research
The AES went through other testing environments, including one in extremely cold temperatures where vaporization can create issues for LiDAR. In wet or dry conditions, the system can move 67.1 cubic meters per hour, similar to the amount of material moved by a human operator in a compact excavator.
"AES performs consistently and reliably over a long time, while the performance of human operators can be uncertain," mentioned Dr. Zhang.
These researchers believe that the new AES shows promise for creating autonomous excavators capable of working in real-world applications. The system has gone through several different testing phases and has managed to operate at similar rates to human operators. The AES aims to decrease human danger and increase safety and automation in the construction industry. Baidu Research Robotics and Auto-Driving Lab and the University of Maryland, College Park plan to continue their research into creating a fully functional autonomous system with refinements to the module in different extreme environments.