Neuroevolution is a powerful method for constructing controllers for nonlinear domains. In this project, the ESP neuroevolution approach is applied to a particular challenging such domain in the real world, controlling a rocket that has no fins. Without fins, the rocket has less drag and flies higher but is unstable. The task for neuroevolution is to learn how to control the amount of thrust in the rocket's four engines to keep it stable. The controller is evolved for the Interorbital Systems, Inc. RSX-2 sounding rocket, using the same simulator, JSBSim, that the industry uses to test new rocket and controller designs. The resulting controller keeps the rocket as stable as the finned version, while flying it 20 miles higher. ESP neuroevolution is therefore an effective method for solving real-world, non-linear control tasks.
See rocket control movie page for a demo.