Learning from Yangtze alligators: A framework for agile and ecologically interactive spine-legged robots

Abstract

A central challenge in bio-robotics is to create machines that can integrate into and illuminate natural ecosystems. The Chinese Yangtze Alligator – a critically endangered species exhibiting exceptionally agile spine-leg coordination honed by its terrestrial-aquatic transition – offers​ a unique model to address this challenge. Yet, existing alligators-like robots fail to capture such biological fidelity due to insufficient actuation, simplified mechanics, and the absence of adaptive control policies. Here, we introduce the Spine-Legged Adversarial Imitation and Reinforcement Learning (SLAIR) framework, which for the first time leverages deep reinforcement learning to master this coordination. By retargeting biological motion data from Yangtze alligators and integrating impedance control to produce natural compliance, our controller achieves adaptive spine-leg coordination in a custom 24-DOFs robot. A variational autoencoder (VAE) generalizes across terrain, while a dual-critic architecture robustly fuses imitation and task rewards. This enables agile locomotion (0.32 m/s, 360° turns in 3.5 s) with a 46.7% reduction in cost of transport. Crucially, the robot’s biomimetic fidelity was validated in the field, where it elicited natural curiosity and approach behavior from wild Yangtze alligators—demonstrating its potential as a transformative tool for conservation biology.

Keywords

Bionic crawling robot / Spine-leg coordination / Adversarial imitation learning / Locomotion control framework / Yangtze alligator

References

[1] P. Ramdya, A.J. Ijspeert, The neuromechanics of animal locomotion: From biology to robotics and back, Sci. Robot. 8 (78) (2023) eadg0279.

[2] G. Jia, et al., Modulating emotional states of rats through a rat-like robot with learned interaction patterns, Nat. Mach. Intell. 6 (12) (2024) 1580-1593.

[3] A.J. Ijspeert, A. Crespi, D. Ryczko, J.-M. Cabelguen, From swimming to walking with a salamander robot driven by a spinal cord model, Science 315 (5817) (2007) 1416-1420.

[4] J.A. Nyakatura, et al., Reverse-engineering the locomotion of a stem amniote, Nature 565 (7739) (2019) 351-355.

[5] K. Melo, T. Horvat, A.J. Ijspeert, Animal robots in the African wilderness: Lessons learned and outlook for field robotics, Sci. Robot. 8 (85) (2023) eadd8662.

[6] X. Zhang, Y. Fang, W. Zhu, X. Guo, A novel locomotion controller based on coordination between leg and spine for a quadruped salamander-like robot, 2019 12th International Workshop on Robot Motion and Control, RoMoCo, IEEE, 2019, pp. 68-73.

[7] Z. Bing, et al., Lateral flexion of a compliant spine improves motor performance in a bioinspired mouse robot, Sci. Robot. 8 (85) (2023) eadg7165.

[8] R. Thandiackal, et al., Emergence of robust self-organized undulatory swimming based on local hydrodynamic force sensing, Sci. Robot. 6 (57) (2021) eabf6354.

[9] K. Karakasiliotis, et al., From cineradiography to biorobots: an approach for designing robots to emulate and study animal locomotion, J. R. Soc. Interface 13 (119) (2016) 20151089.

[10] B. Leung, S. Gorb, P. Manoonpong, Nature’s all-in-one: Multitasking robots inspired by dung beetles, Adv. Sci. 11 (47) (2024) 2408080.

[11] S. Ha, J. Lee, M. van de Panne, Z. Xie, W. Yu, M. Khadiv, Learning-based legged locomotion: State of the art and future perspectives, Int. J. Robot. Res. 44 (8) (2025) 1396-1427.

[12] I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, K. Sreenath, Real-world humanoid locomotion with reinforcement learning, Sci. Robot. 9 (89) (2024) eadi9579.

[13] D. Hoeller, N. Rudin, D. Sako, M. Hutter, Anymal parkour: Learning agile navigation for quadrupedal robots, Sci. Robot. 9 (88) (2024) eadi7566.

[14] X. Cheng, K. Shi, A. Agarwal, D. Pathak, Extreme parkour with legged robots, 2024 IEEE International Conference on Robotics and Automation, ICRA, IEEE, 2024, pp. 11443-11450.

[15] J. Li, S. Huang, X. Xu, G. Zuo, Generative adversarial imitation learning from human behavior with reward shaping, 2022 34th Chinese Control and Decision Conference, CCDC, IEEE, 2022, pp. 6254-6259.

[16] X.B. Peng, P. Abbeel, S. Levine, M. Van de Panne, Deepmimic: Example-guided deep reinforcement learning of physics-based character skills, ACM Trans. Graph. 37 (4) (2018) 1-14.

[17] D. Kang, S. Zimmermann, S. Coros, Animal gaits on quadrupedal robots using motion matching and model-based control, 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS, IEEE, 2021, pp. 8500-8507.

[18] M. Raibert, M. Chepponis, H. Brown, Running on four legs as though they were one, IEEE J. Robot. Autom. 2 (2) (1986) 70-82.

[19] C. Li, M. Vlastelica, S. Blaes, J. Frey, F. Grimminger, G. Martius, Learning agile skills via adversarial imitation of rough partial demonstrations, Conference on Robot Learning, PMLR, 2023, pp. 342-352.

[20] X. Mao, Q. Li, H. Xie, R.Y. Lau, Z. Wang, S. Paul Smolley, Least squares generative adversarial networks, in: Proceedings of the IEEE International Conference on Computer Vision, 2017, pp. 2794-2802.

[21] J. Hwangbo, et al., Learning agile and dynamic motor skills for legged robots, Sci. Robot. 4 (26) (2019) eaau5872.

[22] J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, M. Hutter, Learning quadrupedal locomotion over challenging terrain, Sci. Robot. 5 (47) (2020) eabc5986.

[23] A. Arampatzis, G. Morey-Klapsing, G.-P. Brüggemann, The effect of falling height on muscle activity and foot motion during landings, J. Electromyography Kinesiol. 13 (6) (2003) 533-544.

[24] S. Sood, G. Sun, P. Li, G. Sartoretti, Decap: Decaying action priors for accelerated imitation learning of torque-based legged locomotion policies, 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS, IEEE, 2024, pp. 2809-2815.

[25] Y. Li, L. Zheng, Y. Wang, E. Dong, S. Zhang, Impedance learning-based adaptive force tracking for robot on unknown terrains, IEEE Trans. Robot., (2025).

[26] Y. Tang, L. Hu, Q. Zhang, W. Pan, Reinforcement learning compensated extended Kalman filter for attitude estimation, 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS, IEEE, 2021, pp. 6854-6859.

[27] Q. Shi, X. Huang, B. Meng, Z. Wang, Neural network-based iterative learning control for trajectory tracking of unknown SISO nonlinear systems, Expert Syst. Appl. 232 (2023) 120863.

[28] W. Zhao, J.P. Queralta, T. Westerlund, Sim-to-real transfer in deep reinforcement learning for robotics: a survey, 2020 IEEE Symposium Series on Computational Intelligence, SSCI, IEEE, 2020, pp. 737-744.

[29] N. Rudin, D. Hoeller, P. Reist, M. Hutter, Learning to walk in minutes using massively parallel deep reinforcement learning, Conference on Robot Learning, PMLR, 2022, pp. 91-100.

[30] Q. Wen, et al., Legged robot with tensegrity feature bionic knee joint, Adv. Sci. 12 (12) (2025) 2411351.

[31] N. Kriegeskorte, M. Mur, P.A. Bandettini, Representational similarity analysis-connecting the branches of systems neuroscience, Front. Syst. Neurosci. 2 (2008) 249.

[32] M.H. Dickinson, C.T. Farley, R.J. Full, M. Koehl, R. Kram, S. Lehman, How animals move: an integrative view, Science 288 (5463) (2000) 100-106.

[33] J. Yuan, Z. Wang, Y. Song, Z. Dai, Pitch posture regulation in Peking geckos (Gekko swinhonis): assessing the role of tails before take-off in upward jumping, Biol. J. Linnean Soc. 141 (2) (2024) 238-254.

[34] E. Shahriari, S.A.B. Birjandi, S. Haddadin, Passivity-based adaptive force-impedance control for modular multi-manual object manipulation, IEEE Robot. Autom. Lett. 7 (2) (2022) 2194-2201.

[35] H. Xie, Z. Gao, G. Jia, S. Shimoda, Q. Shi, Learning rat-like behavioral interaction using a small-scale robotic rat, Cyborg Bionic Syst. 4 (2023) 0032.

[36] C. Chen, T.D. Murphey, M.A. MacIver, Tuning movement for sensing in an uncertain world, Elife 9 (2020) e52371.

[37] J. Qiu, et al., A gecko-inspired robot with a flexible spine driven by shape memory alloy springs, Soft Robot. 10 (4) (2023) 713-723.

[38] C. Finn, S. Levine, P. Abbeel, Guided cost learning: Deep inverse optimal control via policy optimization, International Conference on Machine Learning, PMLR, 2016, pp. 49-58.

[39] A. Escontrela, et al., Adversarial motion priors make good substitutes for complex reward functions, 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS, IEEE, 2022, pp. 25-32.

[40] Y. Liu, Z. Liu, Y. Fang, H. Liu, X. Guo, A novel design methodology of CPG model for a salamander-like robot, IEEE Robot. Autom. Lett. 9 (7) (2024) 6115-6122.

[41] W. Haomachai, D. Shao, W. Wang, A. Ji, Z. Dai, P. Manoonpong, Lateral undulation of the bendable body of a gecko-inspired robot for energy-efficient inclined surface climbing, IEEE Robot. Autom. Lett. 6 (4) (2021) 7917-7924.