Optimal Humanoid Ankle Design

• Optimal humanoid ankle mechanism design is focused on replicating human biomechanics through multi-objective optimization and advanced compliant architectures.
• It employs parallel, decoupled, and series compliant mechanisms to achieve natural torque-angle profiles, high agility, and efficient energy recovery.
• Critical methods include finite element and topology optimization to minimize mass while ensuring robust structural integrity and dynamic stability.

Optimal humanoid ankle mechanism design addresses the challenge of replicating or exceeding the functional, kinematic, and energetic properties of the human ankle in robotic or prosthetic systems. The design encompasses kinematics, compliance, actuator architecture, control strategies, and structural optimization, all aimed at generating naturalistic torque-angle and power profiles, high agility and stability, compact mass distribution, and energetic efficiency. Recent research advances iterate on these axes through multi-objective optimization, dynamically-tuned compliance, advanced parallel mechanisms, and detailed benchmarking against human and biological models.

1. Human Ankle Biomechanics and Kinematic Targets

Optimal design strategies are governed by empirical human biomechanics. The human ankle exhibits a cyclic range from approximately –5° plantarflexion at heel-strike to +8° dorsiflexion at mid-stance (≈45% stride), descending rapidly to –25° plantarflexion at push-off. The quasi-static torque profile $\tau_{nat}(\theta)$ peaks near 130 N·m at the end of dorsiflexion, followed by a declining phase during push-off (Baysal et al., 16 Oct 2025). Pulling from these benchmarks, optimal mechanisms must enable:

• Sufficient range of motion (RoM): Dorsiflexion to plantarflexion spanning at least 30–40°, with roll range of ±15° for inversion/eversion (Davydenko et al., 13 Nov 2025).
• Maximal torque output matching or exceeding biological benchmarks: e.g., 130 N·m for high-velocity walking, plus safety margin for agile or perturbed motions (Baysal et al., 16 Oct 2025).
• Compliance tuned to reproduce the natural force-deflection curve, especially for the dorsiflexion "store-and-release" regime critical for gait efficiency (Semasinghe et al., 25 Sep 2025).

Controller architectures must generate variable stiffness, with human-like joint quasi-stiffness ranging 200–450 Nm/rad across stance and push-off phases (Semasinghe et al., 25 Sep 2025). Optimal designs refine this envelope for anthropomorphic energy storage and push-off.

2. Mechanical Architectures: Parallel, Decoupled, and Series Compliant Designs

Ankle mechanism architectures can be summarized under several paradigms:

Parallel Mechanisms (SPU, RSU):

Parallel mechanisms, such as Spherical-Prismatic-Universal (SPU) and Revolute-Spherical-Universal (RSU) architectures, employ multiple legs/rods connecting the tibia and foot. These enable load sharing, improved backdrivability, isotropic manipulability, and compact mass distribution. Both SPU and RSU are modeled as 2-DoF (pitch, roll) devices with optimizable geometry (Cervettini et al., 19 Sep 2025).

Key attributes:

• RSU uses revolute actuators with rotary rods, spherical connectors, and universal joints, with proximal actuator placement for reduced distal inertia.
• SPU features linear actuators for each leg, employing sliders for force transmission, at the cost of higher moving mass.
• Multi-objective optimization tunes link lengths, actuator placement, and joint layout to maximize workspace, isotropy, and dynamic performance (speed, torque), while minimizing mass and backdriving torque.

Decoupled Multi-bar Linkages:

The DecARt leg demonstrates a four-bar multi-link mechanism with actuators above the knee, delivering ankle pitch and roll through distinct rod sets (front/rear), thereby achieving low distal inertia and robust mechanical advantage $M(\theta) > 1$ across the full RoM (Davydenko et al., 13 Nov 2025). This arrangement prevents dead-center singularities, enables smooth transition in load transfer, and directly informs actuator sizing.

Series Elastic and Energy Store†Release Approaches:

Compliant elements (parallel or series springs) are embedded to provide phase-dependent stiffness. The RoboANKLE integrates two parallel linear springs (k₁ = 23 N/mm, k₂ = 16 N/mm) in a dorsiflexion energy-store module and a motorized Extra Energy Storage (EES) unit (K_ES ≈ 45 N/mm) linked via arc sliders, permitting near-complete energy recovery and fine-grained push-off control (Baysal et al., 16 Oct 2025). These mechanisms are optimized to synchronize spring energy release with gait timing, achieving peak torque and power exceeding human reference by 57% and 10% respectively, with prototype mass below 2 kg.

3. Multi-Objective Optimization and Design Frameworks

State-of-the-art design employs multi-objective optimization frameworks that formalize design trade-offs between competing goals: torque, speed, backdrivability, mass, compactness, and stability.

• Variable Geometry: Actuator placement, link lengths, joint types, and solution parameterization (e.g., $\gamma_i$, $\delta_i$ for RSU) are encoded as design variables constrained by kinematic feasibility, actuator limits, and required foot workspace (Cervettini et al., 19 Sep 2025).
• Cost Function Aggregation: Performance is consolidated into a scalar cost ξ, aggregating normalized metrics such as torque capability, speed, backdriving torque, mass, and CoM height with customizable application-based weights.
• Pareto Optimization: Using NSGA-II or similar methods, Pareto-optimal geometries are resolved, followed by scalar scoring across architectures and hardware choices.

Table: Comparison of Optimized Architectures (RSU, SPU, Serial) (Cervettini et al., 19 Sep 2025)

Architecture	Cost ξ (lower=better)	Notes
Serial (original)	0.36	Baseline
RSU (engineered)	0.27	Conventional
SPU (optimized)	0.32	Lower uniformity variance
RSU (optimized)	0.21	Highest agility, lowest mass

The optimized RSU demonstrates a 41% lower cost than serial, with 14% improvement over conventional RSU, indicating that systematic, workspace-feasible parameterization and multi-metric optimization substantially advance ankle module performance.

4. Compliance, Energy Storage, and Bioinspired Push-Off

Compliant and energy-storing elements are essential for both energetic efficiency and dynamic stability:

• Nonlinear Spring Topology: RoboANKLE employs dual-k spring configurations to approximate the nonlinear ankle force-deflection curve. The EES unit, with active motorization, modulates stored energy release timing, providing adaptive torque bursts for push-off while minimizing active motor load (Baysal et al., 16 Oct 2025).
• Bioinspired Tunings: In bioinspired bipedal robots, a monoarticular SOL-equivalent ankle spring (r_SOL ≈ 13 mm, k_SOL ≈ 5–7 kN/m) produces the highest power amplification (up to 5.2x) and lowest cost of transport; biarticular designs favor coordinated ankle-knee actuation for gait stability, but with modest energy gain (Kiss et al., 2022).
• Stiffness Range: Human-inspired joint stiffness profiles, rising from ~200 Nm/rad at touchdown to ~400–440 Nm/rad in mid-stance, can be matched through variable-compliance actuation—whether by hardware tuning or active feed-forward modulation (Semasinghe et al., 25 Sep 2025).
• Energy Storage Targeting: Springs and arcs are dimensioned to store and return ~8 J per step (RoboANKLE); proper pretensioning ensures adequate body support and energetic return at push-off (Baysal et al., 16 Oct 2025).

5. Structural and Topology Optimization

Reducing mass without sacrificing mechanical robustness is critical for ankle mechanisms, especially for robotic prostheses and distal segments.

• Finite Element Analysis (FEA): All critical load-bearing components undergo FEA (e.g., using ANSYS), subject to peak forces and torques derived from dynamic simulation (Adams/MSC). Stress constraints (e.g., Von Mises $\sigma_{max} < 60\%$ yield) and maximum deflection ($\delta_{max} < 2$ mm) are enforced (Baysal et al., 16 Oct 2025).
• Topology Optimization: Objective: minimize volume fraction (e.g., 25%) while retaining structural pathways. Penalization methods (e.g., SIMP, $p=3$) iteratively converge to final geometry, achieving up to 30% mass reduction in foot housing while preserving safety factor $>2$.
• Material Selection: High-stress, low-volume parts leverage composite materials (e.g., Onyx + carbon-fiber), with bulk structures in aerospace-grade Al 7075-T6; shafting uses AISI 302 steel (Baysal et al., 16 Oct 2025).

6. Design Guidelines and Prescriptive Trade-Offs

Experience across platforms yields the following design recommendations:

1. Use series compliance covering 50–70% of peak torque, with nonlinear spring geometry or parallel springs for shaping the torque-angle curve (RoboANKLE, bioinspired bipeds) (Baysal et al., 16 Oct 2025, Kiss et al., 2022).
2. Incorporate a secondary, actively modulated spring for push-off torque augmentation, reducing peak actuator demands by smoothing energy delivery (Baysal et al., 16 Oct 2025).
3. Structure parallel actuation (RSU, SPU) for high backdrivability and proximal motor placement; opt for optimized RSU where mass and stability are critical (Cervettini et al., 19 Sep 2025).
4. Decouple actuator mass from the moving foot—use remote actuation and linkage (multi-bar, cable, rod, or arc-slider) to achieve minimal distal inertia and high agility (DecARt, RoboANKLE) (Davydenko et al., 13 Nov 2025, Baysal et al., 16 Oct 2025).
5. Dimension compliance and rest lengths to provide sufficient pretension at maximal dorsiflexion (≈⅓ body weight) for energy storage and stance support (Kiss et al., 2022).
6. Preserve mechanical advantage $M(\theta) > 1$ across full RoM; avoid singularities and dead-zones via dual-rod or parallel arrangements (Davydenko et al., 13 Nov 2025).
7. Enforce explicit optimization for workspace, speed, torque, backdrivability, and mass; aggregate via normalized scalar cost functions for architecture selection (Cervettini et al., 19 Sep 2025).
8. Total mass should be <2 kg per ankle and envelope dimensions should match the human shank-foot (≈25 cm length) for anthropomorphic fidelity (Baysal et al., 16 Oct 2025).

7. Control, Stability, and Energetic Performance

Control frameworks are coprimary design determinants:

• Forced-Oscillation Templates: Use hybrid controllers matching forced-oscillation axial leg compliance and phase-specific torsional stiffness at the ankle with open-loop or centrally-patterned foot trajectory (Semasinghe et al., 25 Sep 2025).
• Stability Stratification: Large perturbations are rejected by velocity-based foot placement; small deviations are compensated purely via ankle push-off control—demonstrated to sustain stable limit cycles without foot placement correction (Semasinghe et al., 25 Sep 2025).
• Energetics: Empirical validations show optimal mechanisms achieving push-off torque and power outputs exceeding those of biological limbs, with improved energetic margins and cost-of-transport compared to monolithic, actively-controlled designs (Baysal et al., 16 Oct 2025, Kiss et al., 2022).

---

Optimal humanoid ankle mechanisms are realized through the convergence of human biomechanics, multi-objective geometrical optimization, compliant and bioinspired actuation, structural mass minimization, and control frameworks mirroring natural stabilization. Systematic comparative metrics and rigorous simulation-physical validation underpin the progressive refinement of architectures, notably parallel (RSU, SPU), multi-bar linkages, and series-compliant modules, each possessing domain-specific trade-offs in energy, agility, and robustness (Baysal et al., 16 Oct 2025, Davydenko et al., 13 Nov 2025, Semasinghe et al., 25 Sep 2025, Cervettini et al., 19 Sep 2025, Kiss et al., 2022).