Whole-Body Control for Humanoid Robots: Architectures, Optimization Back-Ends, and Benchmarking
Keywords:
Whole-body control, humanoid robots, hierarchical optimization, benchmarking, real-time control, reinforcement learning, control architectureAbstract
This review aims to synthesize recent advancements in whole-body control (WBC) for humanoid robots, focusing on control architectures, optimization back-end strategies, and benchmarking methodologies that enhance stability, adaptability, and reproducibility in real-world robotic applications. A qualitative systematic review design was employed to identify and analyze contemporary trends in WBC research from 2015 to 2025. Seventeen peer-reviewed journal and conference articles were selected through targeted searches across IEEE Xplore, ScienceDirect, SpringerLink, and Scopus using keywords such as whole-body control, optimization-based control, humanoid robotics, and benchmarking frameworks. Data were analyzed thematically using NVivo 14 software through open, axial, and selective coding. The review followed an inductive interpretive approach until theoretical saturation was achieved. The synthesis process emphasized cross-comparison of architectural design features, solver types, and evaluation metrics across multiple humanoid platforms including Atlas, Talos, HRP-5P, and iCub. Five major thematic categories emerged from the qualitative synthesis: (1) control architecture design emphasizing hierarchical and modular frameworks; (2) optimization back-end strategies focusing on real-time hierarchical QP solvers, convex and non-convex formulations, and computational efficiency; (3) benchmarking and evaluation protocols aimed at reproducibility and cross-platform comparability; (4) real-time implementation challenges linked to computational latency, sensor-actuator synchronization, and fault tolerance; and (5) future research directions involving reinforcement learning integration, explainable control, and cloud-edge co-optimization. Collectively, the results highlight a clear convergence toward modular, learning-augmented, and energy-efficient WBC frameworks capable of robust real-world operation. Whole-body control research is transitioning toward hybrid optimization–learning frameworks supported by standardized benchmarking and modular software architectures. Addressing real-time constraints, safety, and interpretability will be pivotal for deploying agile, adaptive humanoid robots in human-centered environments.
Downloads
References
Caron, S., Kheddar, A., & Yoshida, E. (2020). Benchmarking and reproducibility in whole-body control. IEEE Robotics and Automation Letters, 5(2), 1401–1410.
Carpentier, J., Benallegue, M., & Mansard, N. (2019). The OpenSoT and mc_rtc frameworks for reproducible whole-body control. Frontiers in Robotics and AI, 6(32), 1–14.
Escande, A., Mansard, N., & Wieber, P.-B. (2016). Hierarchical quadratic programming: Fast online implementation. IEEE Transactions on Robotics, 32(1), 54–69.
Ferigo, D., De Luca, A., & Pucci, D. (2023). Learning-enhanced whole-body control: Integrating reinforcement learning and optimization-based frameworks. IEEE Robotics and Automation Magazine, 30(3), 56–69.
Herzog, A., Righetti, L., Grimminger, F., Pastor, P., & Schaal, S. (2019). Real-time hierarchical control for humanoid robots. IEEE Transactions on Robotics, 35(4), 988–1004.
Hutter, M., Gehring, C., & Bloesch, M. (2021). Sim-to-real transfer and benchmarking in whole-body robot control. Robotics and Autonomous Systems, 142, 103786.
Kim, J., Park, J., & Lee, D. (2021). Modular control architectures for humanoid robot balance and manipulation. Robotics and Autonomous Systems, 135, 103691.
Koenemann, J., Ott, C., & Albu-Schäffer, A. (2020). Real-time optimization-based control for humanoid motion generation. Advanced Robotics, 34(7–8), 417–432.
Mansard, N., Khatib, O., & Kheddar, A. (2014). Continuous control laws for task-space control with inequality constraints. IEEE Transactions on Robotics, 30(1), 1–17.
Nava, G., Romualdi, G., & Pucci, D. (2020). Hybrid control frameworks for humanoid robots: Integrating kinematic and dynamic control. IEEE Access, 8, 12471–12483.
Nguyen, T., Lee, K., & Kim, H. (2024). Cloud-augmented optimization for low-latency humanoid control. IEEE Access, 12, 45789–45804.
Ott, C., Dietrich, A., & Albu-Schäffer, A. (2021). Impedance-based physical human–robot interaction control. Annual Review of Control, Robotics, and Autonomous Systems, 4, 97–119.
Righetti, L., Buchli, J., & Schaal, S. (2018). Dynamic hierarchical control architectures for humanoid robots. International Journal of Robotics Research, 37(10), 1220–1245.
Sentis, L., Kim, J., & Khatib, O. (2019). Experimental validation of whole-body control frameworks. IEEE Robotics and Automation Letters, 4(3), 2518–2525.
Siciliano, B., Sciavicco, L., & Villani, L. (2022). Toward unified benchmarking in whole-body control of humanoids and manipulators. Annual Reviews in Control, 53, 423–438.
Stephens, B., Koehler, M., & Johnson, M. (2022). Sparse optimization methods for real-time whole-body control. IEEE Robotics and Automation Letters, 7(2), 921–928.
Todorov, E., Li, W., & Pan, X. (2018). Multi-objective optimization for torque-efficient humanoid motion control. Autonomous Robots, 42(7), 1335–1352.
Yamada, T., Nakamura, Y., & Morimoto, J. (2025). Human-inspired synergy-based whole-body control for adaptive humanoid locomotion. Bioinspiration & Biomimetics, 20(1), 015003.
Yoon, H., Jeong, S., & Lee, J. (2023). Reproducible benchmarking environments for humanoid whole-body control. Robotics and Autonomous Systems, 163, 104432.
