Guessing the software arch, you encode the video on the robot and then send it to your backend site. Then you have a separate listener on the pi, that just takes single commands from the website. I imagine not running a webserver on the pi makes the response much faster and doesn't run down the battery. So for a private webcam in my house, a kind of alarm system, i need a system there to capture the video stream and act as a remote host that I can access from outside. I don't want to copy your webserver stack, I'd prefer something lighter weight, that I could run as a service on a linux or windows pc.