The autonomous control of the kite is one of the big issues. It was basically not possible when the idea was first presented in the 80s - there weren't computer

The autonomous control of the kite is one of the big issues. It was basically not possible when the idea was first presented in the 80s - there weren't computers small enough with enough computing power needed for the task. Also there were materials problems. Now we've made a lot of progress on both fronts, but still, the control software is a huge issue. We were planning on using reinforcement learning to learn a controller for our kite (when I was in the biz), but getting enough data to learn on was an issue - the kite design was an iterative process - the kite design would change and that meant we needed to recollect data to enable us to model the new kite changes, wash, rinse, repeat.