As four-legged robots like Boston Dynamics' Spot move into the real world, they must contend with rugged terrain, large climbs and, perhaps most problematically, steep vertical drops. Taking inspiration from a cat's ability to reorient itself in midair, we propose a method to allow robots to reorient themselves during falls, minimizing any fall induced damage. To do so, we simplify the robot as two symmetric cylinders in a zero-gravity environment and used trajectory optimization and reinforcement learning to find a successful trajectory.