Turns out that if you want to stop a looming robot take over of the planet, put a door in front of them. An aerospace engineering professor at the University of Cincinnati (UC), Ou Ma explains, “Robots can do many things, but if you want one to open a door by itself and go through the doorway, that’s a tremendous challenge.”
Doors are such a problem for robots due to the fact they come in many shapes, sizes, colors, and handle heights. The strength needed to open them varies and also most public doors are self closing. Meaning if the robot loses grip it has to start from square one.
Previously studies have tried to scan individual rooms, identify the door and open it. Though this method could only be applied to specific rooms the robot had been trained on. A recent paper from UC, published in IEEE Access, addressed this complicated issue by using machine learning. Using complex AI algorithms in three-dimensional simulations, the team was about to teach a robot to figure out doors through trial and error. After a certain amount of time in training, the machines were able to recognize doors and how to open them with flexibility.
The simulated robot should technically function just like a robot in reality, as the same code is governing their actions. Though to test this out, real life experiments next need to go underway.
Getting over this hurdle is vital for progressing autonomous capabilities. Helper robots are already used in many facilities for vacuuming such as hospitals, airports, and hotels. The next step is further automation, learning how to plug themselves into the wall to charge. When these advances are made, the $27 billion robotics industry can grow even further. This will progress the types of jobs humans no longer have to do, also forwarding industries and economies.
Source study: IEEE – Force-Vision Sensor Fusion Improves Learning-Based Approach for Self-Closing Door Pulling