Few things are more critical to construction on Earth, or offworld, than excavation. Despite this, the difficult to predict nature of digging, and manipulating terrain has standard methods of automation very difficult at meaningful scale.

Until now.

SE4 has a solution, we use human intelligence to designate where to dig and dispose, and allow the automation to complete the repetitive task of scooping up and dumping soil once these two things have been designated.

SE4 has a solution, we use human operators (the best AI there is) to designate dig and disposal locations, then, turn over the repetitive task of the actual digging and disposal to a, just smart enough robotified excavator. It may sound crazy, but once you teach an excavator a library of dig actions, for a suitable library of digable surfaces such work is not at all difficult to automate, due to its repetitive nature. Should there be a specific issue with a surface never before encountered etc, no problem — the system is smart enough to know when it’s outside of its operational ability and is able to call out to its operator for help.

By keeping a human in the loop we’re really leveraging the best of everything, it’s the human operator who has years of experience manipulating many varieties of terrain, in many different weather conditions etc so it’d be foolish not to use that information. The brilliance of the solution is that machines only need to be shown the right way once and then it can be repeated with great fidelity ad nauseam.

When repetitive movements will get the job done, we leave the task to the robot. When you require variety, step in as an operator, let the machine observe you, learn from you and handle it itself next time.

Our model excavator’s edge compute capabilities are powered by NVIDIA’s Xavier modules.

Other projects