actuator selection and integration

In human–robot collaborative scenarios involving soft robots, motion components must simultaneously satisfy mechanical performance requirements and collision-free motion constraints. Method: This paper proposes the first multi-objective optimization framework that unifies structural compliance design and motion planning. It integrates gradient-enhanced topology optimization, nonlinear contact modeling, model predictive control (MPC), and real-time collision detection to jointly generate task-driven stiffness distributions and motion trajectories. Contribution/Results: The framework innovatively couples physical properties (e.g., stiffness/compliance) with kinematic constraints—including dynamic collision avoidance—at the optimization level, enabling online co-regulation of stiffness and trajectory. Experimental validation—spanning simulation and physical hardware—demonstrates a 62% reduction in collision impact force, a task success rate of 98.3%, and an end-to-end response latency under 50 ms.

To address the lack of integrated gearbox parameter optimization and CAD modeling automation in planetary gear actuator design, this paper proposes the first computational framework jointly optimizing gearbox configuration, geometric parameters, and structural layout. The method integrates multi-objective optimization—minimizing mass and axial width while maximizing transmission efficiency—with parametric CAD modeling, enabling fully automated 3D modeling and 3D-printing-ready CAD generation for four planetary gearbox topologies: simple, compound, Woods, and double-simple planetary gearboxes (SSPG, CPG, WPG, DSPG). Its key contributions include systematically characterizing performance boundaries (efficiency, backlash, stiffness) across transmission ratios for each topology and establishing a standardized, motor- and ratio-aware CAD library. Experimental validation shows SSPG achieves 60–80% efficiency, 0.59° backlash, and 242.7 Nm/rad stiffness; CPG attains 60% efficiency, 2.6° backlash, and 201.6 Nm/rad stiffness.

Precision reducers are critical to robotic motion accuracy and dynamic performance, yet challenges persist in contact modeling, stiffness evaluation, and vibration prediction. This work proposes a unified dynamic simulation framework based on explicit contact geometry, integrating advanced contact mechanics theory with efficient numerical solvers to enable rapid reconfiguration across multiple reducer types. The resulting toolkit achieves significantly improved computational efficiency while maintaining high fidelity. Validation against publicly available benchmark data demonstrates that its simulation accuracy for transmission characteristics surpasses that of conventional dynamics software, offering strong generality and extensibility.

Heavy-duty mobile machines (HDMMs) face dual challenges in electrification—constrained by techno-economic limitations—and high-level autonomy—hindered by stringent functional safety requirements. To address these, this project proposes a hierarchical intelligent control framework integrating multibody dynamics modeling, uncertainty suppression, and fault-tolerant mechanisms. Key contributions include: (1) a source-agnostic, modular robust control architecture; and (2) a verifiable AI-control co-design paradigm, enabling formal co-verification of deep reinforcement learning policies with ISO 13849 functional safety standards. The approach combines nonlinear robust control, adaptive observers, policy distillation, and multiphysics co-simulation, validated via hardware-in-the-loop testing across three HDMM platforms. Results demonstrate significant improvements in system response robustness and fault recovery capability. Outcomes are disseminated in five peer-reviewed publications and support industry-wide safe autonomous upgrading.

Design optimization of flexible multibody systems—comprising elastic or deformable components—faces significant challenges in accurate, efficient gradient computation for full dynamic sensitivity analysis. Method: This paper proposes an efficient gradient computation framework based on the discrete adjoint variable method, systematically applied to comprehensive dynamic sensitivity analysis for the first time. Integrating the Absolute Nodal Coordinate Formulation (ANCF) with nonlinear dynamic modeling, the framework enables concurrent optimization of structural parameters and control strategies. Contribution/Results: Unlike conventional finite-difference and continuous adjoint approaches, the proposed method eliminates truncation errors and numerical instabilities, achieving analytical-grade accuracy (gradient error ≤ 1×10⁻⁶) and high computational efficiency. In case studies involving spacecraft deployment mechanisms and high-speed robotic manipulators, optimization convergence accelerates by over fivefold, substantially advancing the practical implementation of integrated topology–sizing–control optimization for large-scale flexible systems.

To address the high complexity and mass associated with conventional robotic variable-transmission systems—typically reliant on auxiliary actuators—this paper proposes a load-based passive variable-transmission (LBVT) mechanism. The LBVT integrates preloaded springs with a four-bar linkage, enabling purely mechanical, load-responsive adaptation of the gear ratio: it automatically increases the transmission ratio when joint load exceeds 18 N, thereby enhancing torque output. Crucially, it operates without sensors or active control, departing from the traditional closed-loop control paradigm for variable transmission. Theoretical modeling and parametric simulation demonstrate that the transmission ratio can increase by up to 40%, significantly improving actuation efficiency and robustness under dynamic loading conditions—such as those encountered in legged robots—while simultaneously reducing system complexity and mass.

This study addresses the challenge of simultaneously achieving component alignment, system coordination, solution reliability, and computational efficiency in physically interacting interconnected systems within three-dimensional space. To this end, the authors propose a decomposition-based collaborative optimization framework that, for the first time, embeds port-alignment constraints into the SPI² architecture. Treating component positions as design variables, the method employs a penalty function to enforce system-level feasibility and enables automatic generation of initial designs. By integrating gradient-based optimization for enhanced numerical stability and coupling it with NSGA-II for efficient multi-objective search, the approach achieves high-quality coordinated solutions. Demonstrated on automotive powertrain and battery-chassis integration cases, the framework significantly outperforms discrete exhaustive search, delivering superior system-level coordination while substantially reducing computational cost.

This work addresses the limitations of existing co-design approaches for legged robots, which often neglect actuator parameter optimization and are confined to open-chain architectures, thereby hindering high-performance jumping. The paper proposes a co-design framework tailored for a planar closed-chain five-bar monopedal robot, uniquely incorporating detailed actuator specifications—such as motor and gearbox mass, efficiency, and peak torque—into whole-system optimization. The framework jointly optimizes mechanical structure, actuator selection, and control policy through a two-stage methodology: first establishing a mapping between gear ratios and actuator performance, then applying the CMA-ES algorithm for global co-optimization. Simulation results demonstrate that the optimized design achieves a 42% increase in jump distance and a 15.8% reduction in mechanical energy consumption compared to a baseline configuration.

Concurrent lightweighting and durability enhancement of compliant-link manipulators remains challenging due to the intrinsic trade-offs among vibration suppression, structural mass, and fatigue life under coupled compliance-dynamics effects. Method: This paper proposes a multi-objective geometric sizing optimization framework incorporating fatigue-life constraints. It integrates a rainflow-counting–based fatigue estimation model with the critical plane method, employing Tresca equivalent stress and linear damage accumulation. The framework co-optimizes trajectory planning (time- and energy-optimal) and dynamic simulation to simultaneously minimize mass, suppress vibration, and maximize fatigue life on the Pareto front. Contribution/Results: Validated on a 3-DOF serial manipulator performing pick-and-place tasks, the approach reduces vibration amplitude significantly, improves fatigue life by 32%, and decreases mass by 18%. This work is the first to incorporate the critical plane method into the structure–control co-design of flexible manipulators, establishing a scalable, durability-driven design framework for elastic mechanisms.

This work addresses the dual challenges of actuator configuration optimization and sensorless control for fully electric-driven heavy-duty manipulators. Method: We propose an integrated modeling–optimization–control framework: (i) a mechatronically coupled dynamic model; (ii) a physics-informed Kriging surrogate model incorporating both physical constraints and data-driven learning; (iii) joint optimization of actuator parameters and dual-variable (force/velocity) sensorless estimation via NSGA-II multi-objective optimization combined with a lightweight neural network; and (iv) hierarchical virtual decomposition control to ensure high-fidelity trajectory tracking. Results: Experiments under variable-load conditions demonstrate high-precision trajectory tracking (RMSE < 0.8°), robust sensorless force/velocity control, 19.3% energy efficiency improvement, and 22.7% reduction in peak power consumption—significantly enhancing operational safety and engineering deployability.