Tumbling robots are simple platforms that are able to traverse large obstacles relative to their size, at the cost of being difficult to control. Existing control methods apply only a subset of possible robot motions and make the assumption of flat terrain. Reinforcement learning (RL) allows for the development of sophisticated control schemes that can adapt to diverse environments. By utilizing domain randomization while training in simulation, a robust control policy can be learned that transfers well to the real world. In this article, we implement autonomous set point navigation on a tumbling robot prototype and evaluate it on flat, uneven, and valley–hill terrain. Our results demonstrate that RL-based control policies can generalize well to challenging environments that were not encountered during training. The flexibility of our system demonstrates the viability of nontraditional robots for navigational tasks.