Contact Us
Developing AI to Realize Automation (Autopilot) of Construction Machinery
Project Summary
BY APPLYING AI/REINFORCEMENT LEARNING AND IMITATION LEARNING TECHNOLOGIES TO AUTOMATE THE OPERATION OF CONSTRUCTION MACHINERY ("CONSTRUCTION EQUIPMENT") AND OTHER EQUIPMENT, ARAYA'S AUTONOMOUS AGENT TEAM AIMS TO SIGNIFICANTLY IMPROVE THE EFFICIENCY OF OPERATIONS.
Social context of the project
As the working population shrinks due to the declining birthrate, there is a need to improve the efficiency of construction work. In addition, as the number of skilled technicians handling construction equipment ages, there is a growing need to pass on skills.
To address these issues, ARAYA will apply reinforcement learning and imitation learning technologies, which are part of AI technology, with the aim of realizing autonomous construction equipment that can work autonomously and efficiently.
Technology to realize "Autonomous Construction Machinery
There are two methods to create autonomous construction equipment:
The first is a method that uses "reinforcement learning". The autonomous construction machine learns by trial and error of various maneuvers in its own environment (simulation environment or actual site). It takes a lot of time for the machine to learn, but there is a possibility that the machine can acquire more control than a human.
The second method is based on "imitation learning". Autonomous construction machinery learns equivalent operations based on model data from humans (skilled technicians). It is thought to be possible to achieve the same level of movement as humans in a relatively short time.
Technology to realize "Autonomous Construction Machinery
Since a great deal of trial and error is required in the above learning process, it is not realistic to try it on the actual machine from the beginning in terms of cost and time. In order to solve this problem, ARAYA utilizes Digital Twin to evaluate and examine the feasibility and validity of the design by repeating the trial and error process on a simulator. After that, we make adjustments so that the learned model will work properly with actual construction equipment and sites.
Technology to realize "Autonomous Construction Machinery
Since a great deal of trial and error is required in the above learning process, it is not realistic to try it on the actual machine from the beginning in terms of cost and time. In order to solve this problem, ARAYA utilizes Digital Twin to evaluate and examine the feasibility and validity of the design by repeating the trial and error process on a simulator. After that, we make adjustments so that the learned model will work properly with actual construction equipment and sites.
Benefits of Reinforcement Learning and Imitation Learning for Construction Equipment Autonomy
We believe that the autonomous construction equipment developed by ARAYA will have the following advantages.
Example of current results 1: Excavation of soil by hydraulic excavator
The industrial simulator " Vortex Studio " (provided by CM Labs Simulations and Dentsu International Information Services) was used to train the hydraulic excavator to excavate a certain amount of soil.
At first, it couldn't scoop any soil at all, but it gradually learned the angle of the bucket and when to move it, and was able to scoop more than a certain amount of soil.
Other examples of the use of construction equipment, etc.
In addition to more difficult tasks with hydraulic excavators and cranes, the following construction equipment may be used.
Embankment and land leveling by bulldozers
Loading and unloading of containers by gantry cranes
Loading, unloading, and transporting cargo with forklifts
For the future
CURRENTLY, WE ARE DEVELOPING SOME CONSTRUCTION EQUIPMENT AND TASKS IN A SIMULATION ENVIRONMENT, BUT IN THE FUTURE, IT WILL BE IMPORTANT TO DEVELOP CONSTRUCTION EQUIPMENT AND TASKS THAT HAVE HIGHER NEEDS IN THE FIELD, AND TO DEVELOP THEM THROUGH EXPERIMENTS IN A REAL ENVIRONMENT. WE ARE LOOKING FOR PARTNER COMPANIES IN THE CONSTRUCTION INDUSTRY, ETC., WHO ARE WILLING TO DEVELOP TOGETHER WITH US BY UTILIZING ARAYA'S AI, REINFORCEMENT LEARNING, AND IMITATION LEARNING TECHNOLOGIES.
