Abstract: Tumble stability is critical to a mobile robot cruising in uneven terrains.A wheel-legged robot with symmetrical structure is developed,which has four independent wheel-legged articulations and is able to generate a series of configurations. A dynamic energy stability pyramid and a tumble stability index are adopted to synthetically evaluate stability of the robot.An adaptive control system using fuzzy neural network method is created.With the tumble stability index computed, the robot configuration and velocity can be changed in real time to guarantee its tumble stability.Simulation results on a sinusoidal road surface show that the behaviors yielded by the system are of high real-time performance and reliability,and the system can reduce risks in robot autonomous negotiation.