Upon an unexpected launch, a squirrel demonstrates the ability to control its orientation while its body tumbling in air. It can turn toward and then lock its head (and thus eyes) towards a fixed spot (usually the landing spot) until it safely landed on feet. Such ability of squirrels could serve as inspiration for developing bio-inspired solutions to address challenging landing problems for multibody systems in unstructured environments. In this paper, we first proposed a 3D multibody dynamics model to simulate the motions of a squirrel-like quadruped robot. Building on our prior research on squirrel-like planar orientation control, we designed rewards in reinforcement learning (RL) algorithm (Deep Deterministic Policy Gradient, DDPG) to train the agent to align its head toward a predetermined landing target. This approach was then applied to plan joint motions while the robot’s body tumbling and falling in the air for safe landing.