Hitachi Construction Machinery is set to begin research and development in October to train artificial intelligence to autonomously operate an excavator with the ability to adapt in real time it to its environment and conditions.
The goal of the project, to be conducted in Japan, is to achieve autonomous operation for disaster debris removal and for excavation and loading work at construction sites and mines, according to Hitachi.
The company is collaborating with Tokyo-based Jizai Inc. and the Nara Institute of Science and Technology. The project has received the backing of the Japanese government, which seeks to strengthen the country's generative AI development. Generative AI uses data to train AI models for real-world applications.
In Hitachi’s case, the company says it “will collect extensive operator work footage and operation logs to train the physical AI under development.” Physical AI refers to a machine’s ability to interact in the physical world.
The physical AI will be trained to make decisions and perform tasks on actual construction sites by feeding it operational data from a variety of conditions. Such training would not only cover regular operations but also “recovery processes from operational errors, weather, the surrounding environment, machinery conditions, object positioning, etc.,” according to Hitachi.
Powered by“The project aims to realize construction machinery that autonomously performs tasks according to the site environment by developing a new AI trained on the operational data of human operators.”
The catalyst for the project is the growing labor shortage, as well as to improve safety. The company also notes the need for safe debris removal from disaster scenes, where it can be dangerous to send in workers.
Hitachi says the project could lead to future development of physical AI in other fields, including material handling, forestry, demolition and civil engineering.
