Baidu AI Researchers Introduce An Autonomous Excavator System (AES) To Perform Material Loading Tasks Without Any Human Intervention

2500
Source: http://research.baidu.com/Blog/index-view?id=159

Baidu Research Robotics and RAL (Auto-Driving Lab), along with the University of Maryland, College Park, have introduced a new benchmark in robotics. An autonomous excavator system (AES) has been manufactured that is claimed to give excellent results. The AES would be able to:

  • Work efficiently without any form of human intervention.
  • Would not compromise on the quality of services; that is, the performance levels have been found to mimic those of an experienced human worker closely.
  • It can work for more than 24 hours and has been tested for the same in real-world scenarios (more than ten different scenarios have been tested to understand the working properly).
  • It increases the productivity of the work manifold without any hazards attached.

If the autonomous excavator is as successful as is being claimed, it could be a new direction in robotics, especially for using them for long durations. As of now, Baidu has already started to collaborate with numerous of the leading construction companies all around the globe to automate the traditional heavy construction machinery.

Excavations are an essential part of the industry and necessary for construction as well as mining activities. Currently, the global market size for excavators was found to be around $44.12 billion in the year 2018, but it is expected to grow and come around the figure of $63.14 billion. This widespread expansion would lead to the demand for an unprecedented amount of skilled machinery workers, in which case robots could come in extremely handy. Moreover, the environment for mining and construction activities has usually been found as toxic, which again leads to an apprehension among humans to join this field and an additional headache for the employers. Brining in robots to the same job would be a rational decision, one that would be backed with a plethora of scientific reasons. The coming of robots such as AES could prove to be an impetus to other fields of manufacturing, warehousing, and autonomous vehicles, which could be ground-breaking for markets worldwide.

The AES is a different form of robot altogether since mostly the robots used in the industries are small and are only equipped to function in particular environments. On the other hand, the excavating robots have to be agile and versatile to identify the target materials, avoid all obstacles, and handle the varying environmental conditions. For the same reason, the AES has been powered with accurate and real-time algorithms for proper planning, perception, and reasonable control. A new architecture has been built with multiple sensors, including the LiDAR cameras and the proprioceptive sensors, to include all of these properties. These combined give a 3D environment to the robots, and the advanced algorithms navigate the neural networks to provide clean images.

In order to properly assess the working of the AES, the researchers also worked in collaboration with a leading manufacturing enterprise to deploy the robot at different scenes and test its efficiency, especially in the arenas of waste management wherein robust automation is the demand of the day.

The researchers’ focus is to further refine the AES for even better performance, especially in extreme weather conditions. The robot has already been tested in winters for the potential vaporization that can affect the normal functioning of the robot, especially the LiDAR. The results claimed have been good in both wet and dry conditions. The compact excavator was able to extract 67.1 cubic meter that is similar to human performance. The main advantage, however, is the reliability and the consistency of the working of the AES.

Source: http://research.baidu.com/Blog/index-view?id=159

Paper: https://robotics.sciencemag.org/content/6/55/eabc3164

https://www.youtube.com/watch?v=KFcNf_k0E_M