Submitted by schulzb5 on Thu, 2018-03-08 07:44
Recent advances have been made in the realm of machine learning through simulation. One advance was Nvidia releasing the Jetson platform and the Isaac virtual robot. Through iterative learning virtual robotic arms in a simulated environment were able to properly grasp all sorts of objects. This learning was able to be transferred to a real world robotic arm. It would be interesting if Tesla was able to take that tech and use it towards letting the robotic arms program themselves when it came to optimum movements, most efficient/strong/cost effective/timely /alloy type of welds. Simulating million of permutations might alleviate some the growing pains the gigafactory is currently having.