This article explores the emergence of gait transitions in agile quadrupedal locomotion on challenging terrains, focusing on the role of viability (avoiding falls) as a key criterion. The study uses a hierarchical biology-inspired framework combining robotics and deep reinforcement learning (DRL) to investigate the interaction between supraspinal drive, the central pattern generator (CPG) in the spinal cord, the body, and exteroceptive sensing. Key findings include: 1. **Viability as a Primary Criterion**: The transition from walking to trotting improves both viability and energy efficiency, suggesting that viability is a primary objective for gait transitions. 2. **Discrete Terrain Challenges**: On discrete gap terrains, the emergence of trot-pronk transitions helps avoid non-viable states, further supporting the role of viability. 3. **Robot Performance**: The proposed framework demonstrates state-of-the-art agility in challenging scenarios, outperforming existing controllers in gap-crossing tasks. 4. **Sensory Information**: Exteroceptive sensory information, particularly distances to gaps, is crucial for successful gait transitions and gap crossing. The study provides insights into the underlying mechanisms of gait transitions and highlights the potential of using robots to test biological hypotheses about animal locomotion. Future work could explore more detailed models of the musculoskeletal system and incorporate additional sensory feedback to enhance the understanding of gait transitions.This article explores the emergence of gait transitions in agile quadrupedal locomotion on challenging terrains, focusing on the role of viability (avoiding falls) as a key criterion. The study uses a hierarchical biology-inspired framework combining robotics and deep reinforcement learning (DRL) to investigate the interaction between supraspinal drive, the central pattern generator (CPG) in the spinal cord, the body, and exteroceptive sensing. Key findings include: 1. **Viability as a Primary Criterion**: The transition from walking to trotting improves both viability and energy efficiency, suggesting that viability is a primary objective for gait transitions. 2. **Discrete Terrain Challenges**: On discrete gap terrains, the emergence of trot-pronk transitions helps avoid non-viable states, further supporting the role of viability. 3. **Robot Performance**: The proposed framework demonstrates state-of-the-art agility in challenging scenarios, outperforming existing controllers in gap-crossing tasks. 4. **Sensory Information**: Exteroceptive sensory information, particularly distances to gaps, is crucial for successful gait transitions and gap crossing. The study provides insights into the underlying mechanisms of gait transitions and highlights the potential of using robots to test biological hypotheses about animal locomotion. Future work could explore more detailed models of the musculoskeletal system and incorporate additional sensory feedback to enhance the understanding of gait transitions.