The mathematical core: how WBC structures the control problem
Whole-body control treats the humanoid robot as a single kinematically and dynamically coupled system rather than decomposing it into independent subsystems. The framework defines a compound task consisting of operational-space objectives — center-of-mass trajectory, foot placement, end-effector pose — alongside lower-priority goal postures, and resolves all tasks simultaneously against a constraint set encoding contact forces, joint torque bounds, and friction cone membership. This architecture was formally defined by the Board of Regents of the University of Texas System in a 2016 patent, which specifies a robot model that computes kinematic and dynamic properties, binds controller parameters to transport layers for external process access, and instantiates the WBC using the compound task and constraint definitions.
The mathematical substrate of WBC is inverse dynamics in the operational space. Joint accelerations are computed by mapping Cartesian-space acceleration commands through the robot’s forward kinematics Jacobian, and ground reaction forces at multiple contact points are optimized to satisfy Newton-Euler equations of motion. A 2018 filing from Shenzhen Weiteshi Technology describes the whole-body dynamic controller operating in the operational space as an acceleration command computation, converting commands to joint accelerations via differential forward kinematics, and optimizing contact reaction forces for under-actuated robot dynamics — with support for simultaneous computation of position- and time-parameterized outputs enabling generation of multiple walking patterns at real-time speeds.
The SRBD model is a common simplification adopted within WBC pipelines to enable real-time execution. Leg links are treated as massless connectors and total body mass is concentrated at the floating base. Translational dynamics follow Newton’s second law across a finite set of contact points, while rotational dynamics follow the Euler equation with the Coriolis/centrifugal term neglected at low speeds — yielding a compact state-space representation amenable to both MPC and learning-based optimization, as demonstrated on the Unitree H1 humanoid by Electronic Science and Technology University Zhongshan Campus (2025).
Priority-ordered task hierarchies are central to WBC’s ability to simultaneously satisfy multiple conflicting objectives. A 2025 filing from Beijing Institute of Technology describes a tracking control module that employs a full rigid-body dynamics model, resolves joint torques through inverse dynamics, implements motion tracking via WBC, and uses hierarchical optimization to prioritize motion tasks over contact force tasks — a canonical formulation repeated across the corpus. The state estimation module supplies all current state quantities required by both the upstream MPC and downstream WBC layers, illustrating the standard two-level architecture.
Whole-body control (WBC) in humanoid robots formulates locomotion as a prioritized constrained optimization problem — typically a quadratic program — that simultaneously resolves center-of-mass trajectory, foot placement, end-effector pose, joint torque bounds, and friction cone constraints across all degrees of freedom in a single computation.
MPC + WBC: the two-layer architecture that dominates practice
The dominant engineering solution for model-based humanoid locomotion pairs WBC with model predictive control in a two-layer hierarchy: MPC handles trajectory generation over a receding horizon, while WBC translates the resulting reference trajectories into feasible joint torques accounting for the robot’s full dynamic model at the current time step. This separation of concerns — long-horizon planning versus instantaneous torque resolution — is described across multiple 2025 filings and represents the current state of practice.
Huazhong University of Science and Technology’s 2025 filing introduces a full-body kinematics–centroidal dynamics model that generates complete walking trajectories including upper-body joint motion in real time. The MPC model unifies trajectory generation and online tracking control, with optimal state and control input passed as inputs to the WBC layer. The centroidal dynamics formulation explicitly includes linear momentum and angular momentum at the center of mass, contact forces at each contact point, and the cross-product moment arm from the CoM to each contact site — enabling arm and waist dynamics to contribute to whole-body balance rather than being treated as passive appendages.
“Prior WBC approaches fail to include upper-body joint motion in the planning stage, leaving WBC to compensate inadequately after the fact — this work achieves genuinely whole-body coordinated walking rather than locomotion with appended arm motion.”
Zhejiang University’s 2025 centroidal dynamics patent constructs a nonlinear model predictive control (NMPC) problem that simultaneously plans CoM state, arm end-effector state, foot state, and control inputs. Equality and inequality constraints governing whole-body coordinated motion are designed into the optimization, including task trajectory constraints on arm end-effectors and momentum conservation constraints that leverage arm motion to assist locomotion balance. An inverse dynamics layer then computes feed-forward joint torques, combined with joint-level PD controllers for closed-loop tracking.
In the standard MPC+WBC two-layer architecture for humanoid robots, model predictive control plans contact sequences and center-of-mass trajectories over a receding horizon, while whole-body control resolves those trajectories into per-joint torques at the current time step using the robot’s full dynamic model — a separation described by Beijing Institute of Technology (2025) and Zhejiang University (2025).
Tsinghua University’s 2024 filing introduces collision-aware WBC (CA-WBC) for legged robots, implemented as a weighted quadratic program (WQP). When a stance leg unexpectedly fails to contact the ground — a “late collision” scenario — the WBC switches objectives and constraints according to a decision tree, adding discrete collision model terms to the optimization. The authors explicitly identify that computational resources on real robots are constrained, motivating the WQP implementation over more expensive second-order cone program (SOCP) formulations.
Beijing Xiaomi Robot Technology’s 2024 filing applies offline discretization of full dynamic model parameters indexed by key joint angles, using grid interpolation at runtime to recover per-joint dynamics parameters and generate WBC joint torque commands — explicitly reducing computational load while maintaining full-dynamics accuracy. Yibajwu Robotics (Suzhou) describes a 2026 hierarchical dynamics optimization model that fuses real-time joint signals with historical optimization data, running a motion simulation layer to load environmental constraint data and refine joint torque allocation via a cross-joint torque coupling optimization unit, with feedback weights updated continuously based on deviation tracing information.
Explore the full patent landscape for whole-body control and humanoid robot locomotion in PatSnap Eureka.
Search WBC Patents in PatSnap Eureka →Reinforcement learning augments — and sometimes replaces — classical WBC
Classical WBC requires accurate dynamic models and struggles with robustness on unstructured terrain. The corpus documents a strong trend toward incorporating reinforcement learning to address these limitations — either by learning policies that output targets consumed by downstream WBC layers, or by replacing portions of the WBC stack with learned neural networks. This trend is explicitly motivated in a 2025 iFLYTEK patent that catalogs Zero Moment Point (ZMP), Model Predictive Control (MPC), and Whole Body Control (WBC) as classical locomotion algorithms, then identifies their shared limitation: mathematical model simplifications that yield stable behavior only in specific environments and fail on unstructured terrain.
Hangzhou Dianzi University’s 2026 filing addresses a specific computational bottleneck: centroidal momentum is traditionally computed on CPU using libraries such as Pinocchio, creating a data transfer bottleneck when RL training runs on GPU. The authors construct CasADi symbolic expressions for joint-space inertia tensors and centroidal momentum, compile these to CUDA kernels, and execute them in parallel across thousands of simulation environments on GPU — enabling reward functions that supervise whole-body coordination rather than just joint positions or CoM velocity.
Electronic Science and Technology University Zhongshan Campus (2025) trains a fully-connected residual network offline to approximate MPC-optimal ground reaction forces from CoM state inputs, with a priority-based loss function that enforces friction cone and reaction force bound constraints. During online deployment, the network replaces the MPC solver, providing efficient inference. The residual network outputs are fused with joint angles, angular velocities, and prior actions, then processed through a reinforcement learning policy trained with biomechanically-referenced trajectories from public human motion datasets. A symmetric loss constraint on lower-limb motion improves convergence speed, and the final policy constitutes a whole-body control strategy incorporating both model-based prior knowledge and human motion references.
For biped robots specifically, Zhejiang University of Technology’s 2026 filing demonstrates that running and jumping can be learned without complex demonstration data by designing reference trajectories for leg joint pitch angles and body pose, then using RL with a two-stage training protocol — base policy first, then strict optimization — with Proximal Policy Optimization (PPO) and asymmetric Actor-Critic networks. Domain randomization enhances sim-to-real transfer stability. According to IEEE Robotics and Automation, sim-to-real transfer remains one of the central open challenges in legged locomotion research.
Hangzhou Dianzi University’s 2026 patent on whole-body control reinforcement learning for humanoid robots compiles centroidal momentum computations to CUDA kernels and executes them across thousands of parallel GPU simulation environments, enabling RL reward functions that directly supervise whole-body coordination rather than only joint positions or center-of-mass velocity.
The 2018 Shenzhen Weiteshi Technology filing proposes an integrated framework combining a phase-space planning (PSP) framework, a robust reinforcement learning process exploiting PSP’s inherent directional walking constraints, and a whole-body dynamic controller. The RL process operates over simplified models derived from PSP, while the WBC layer handles full-body dynamics in the operational space — generating multiple walking patterns at real-time computational speed and addressing the longstanding gap in 3D full-body humanoid dynamic walking under RL guidance.
Analyse reinforcement learning and WBC patent filings across jurisdictions with PatSnap Eureka’s AI-powered search.
Explore RL + WBC Patents in PatSnap Eureka →Arms, waist, and upper-body: the full-body coordination challenge
A distinguishing challenge of humanoid WBC versus quadruped WBC is the need to coordinate upper-body degrees of freedom — arms, waist, and torso — with lower-body locomotion. Multiple 2025 patents explicitly address this problem, and the consensus finding is that upper-body dynamics must be incorporated at the planning stage, not left to WBC post-hoc correction, to achieve genuinely coordinated whole-body motion.
Zhejiang University’s NMPC formulation (2025) includes momentum conservation constraints that leverage arm motion to assist locomotion balance, designing these constraints directly into the optimization problem rather than treating arm motion as a secondary concern. The patent demonstrates that arm–body coordination, historically treated as peripheral, is now being systematically incorporated into the WBC planning problem — a shift validated independently by Huazhong University of Science and Technology’s full-body kinematics–centroidal dynamics model, which includes nc contact points with three-dimensional contact forces and full centroidal momentum (linear plus angular) as state variables.
Zhiyuan Innovation (Shanghai) Technology’s 2025 filing addresses the specific challenge that vigorous upper-limb motions destabilize the lower body. A prediction model trained on dynamically-augmented upper-limb trajectory samples learns to map upper-limb motion information and robot state to lower-limb joint angles, ensuring that even in cases of violent arm motion the lower body can maintain balance. Dynamic augmentation during training varies joint masses, control coefficients, angles, and velocities, creating a robust learned correspondence between upper and lower body dynamics.
Zhejiang University’s 2025 patent on whole-body coordinated control for humanoid robots incorporates momentum conservation constraints that leverage arm motion to assist locomotion balance directly within the nonlinear MPC optimization, rather than leaving arm–body coordination to the WBC layer as a post-hoc correction.
Xiaopeng Motors’ 2025 filing introduces a hybrid full-body control approach that separately governs the lower limbs and waist — via torque commands from an RL model masked to the relevant joints — and the upper limbs — via joint position commands derived from reference motion data — then concatenates both command sets into a unified whole-body control instruction. This hybrid strategy explicitly separates the locomotion-stability-critical lower body from the manipulation-oriented upper body while maintaining a coherent whole-body command interface.
Tencent’s 2025 filing extends WBC beyond pure locomotion to non-prehensile manipulation on a wheel-legged platform, incorporating passivity-based control (IDA-PBC) for a nonlinear sphere-balancing task into the whole-body control architecture. The base control information for maintaining sphere balance is computed via passive control of the nonlinear coupled robot-sphere system, then fed into the WBC layer that determines per-joint torques — demonstrating WBC’s extensibility to loco-manipulation tasks requiring simultaneous balance and object control. This aligns with research directions highlighted by Science Robotics on unified loco-manipulation frameworks for humanoid platforms.
Key patent holders and innovation clusters shaping WBC
Based on frequency and technical depth of relevant filings across the corpus of more than 50 patents, the following assignees constitute the dominant innovation clusters in whole-body control for humanoid and legged robots.
Tencent Technology (Shenzhen)
Tencent is the most prolific assignee in the corpus with at least six distinct active patents covering dynamic whole-body dynamics control, adaptive controller construction via policy and value iteration, WBC integration with passivity-based manipulation, and sim-to-real policy transfer with robustified state representations — spanning WO, CN, and EP jurisdictions. According to WIPO patent data, cross-jurisdictional coverage of this breadth signals strategic commercial intent across major robotics markets.
Zhejiang University & Huazhong University of Science and Technology
Zhejiang University contributes foundational work on centroidal dynamics-based NMPC for whole-body coordination (2025), broadening its portfolio across robot morphologies. Huazhong University of Science and Technology focuses on online generation of full-body coordinated walking trajectories that integrate upper-body dynamics from the planning stage (2025) — representing a theoretical advance over MPC+WBC pipelines that treat arm motion as an afterthought.
Electronic Science and Technology University Zhongshan Campus
This institution addresses the practical problem of replacing online MPC solvers with learned approximators for real-time WBC execution (2025), using a residual network trained with biomechanically-referenced human motion data and a symmetric loss constraint on lower-limb motion to improve convergence.
University of Texas System (Board of Regents)
The UT System holds a foundational WO patent (2016) that defines the software architecture for binding WBC to transport layers and external processes — a framework-level contribution underpinning many subsequent implementations. The PatSnap patent analytics platform identifies this filing as a high-citation anchor in the WBC citation network.
Hangzhou Dianzi University & Beijing Institute of Technology
Hangzhou Dianzi University addresses the GPU-CPU bottleneck in RL-based WBC training via CUDA-accelerated centroidal dynamics (2026) — a practical systems contribution enabling higher-fidelity whole-body reward functions in large-scale RL training. Beijing Institute of Technology contributes slip-aware WBC for wheel-legged robots (2025), extending WBC to hybrid wheel-leg platforms operating on slippery or irregular terrain.
The PatSnap innovation intelligence platform tracks more than 2 billion data points across 120+ countries, enabling IP teams to monitor emerging WBC filings as they publish across these jurisdictions in real time.