One-DoF Robotic Design of Overconstrained Limbs with Energy-Efficient, Self-Collision-Free Motion

Jia Pan, Yuping Gu, Bangchao Huang, Haoran Sun, Ronghan Xu, Jiayi Yin, Wei Zhang, Fang Wan, Jia Pan 0001, Chaoyang SongarXiv 2025 · DOI · arXiv

Introduction

Robotic limb design often faces a trade-off between the flexibility of multi-degree-of-freedom (DoF) systems and the simplicity of 1-DoF mechanisms. While 1-DoF limbs are lighter and more robust, they are typically restricted to simple planar motions and frequently encounter self-collision issues. This paper presents a computational framework for designing overconstrained linkages, such as the Bennett linkage, to follow complex 3D trajectories while ensuring energy efficiency and collision-free operation.

Methods

The proposed framework utilizes a two-stage optimization pipeline to bridge the gap between abstract kinematic models and physical robotic hardware.

Kinematic & Parametric Optimization

The first stage identifies optimal linkage parameters—such as link lengths and twist angles—to match a target 3D trajectory (e.g., an insect's walking gait). A key innovation is the Modified Spatial Curve Similarity Metric, which is invariant to translation, rotation, and scaling, making it robust for matching complex 3D paths. The objective function also incorporates simulated torque and mechanical energy to ensure the synthesized motion is energetically efficient, using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) for optimization.

Geometric Optimization for Collision Avoidance

Once the kinematic skeleton is defined, the physical shape of the links is optimized. Links are modeled as curved bars parameterized by polynomials. To prevent self-collision, the authors introduce a Continuous Collision Energy Function.

Kinematically equivalent forms and collision avoidance.

By representing each link as a chain of particles, a smooth, differentiable energy term is formulated. This term "pushes" links apart during optimization without altering the underlying kinematics. This decoupling allows the design to achieve a full 360-degree rotation in 3D mechanisms that would otherwise lock up due to physical interference.

Physical link implementation and the collision energy function.

Key Insights

The central contribution of this work is the decoupling of a linkage's kinematic parameters from its geometric shape. By curving the links, the researchers can realize complex 3D motions with a single motor. Furthermore, dynamic-aware synthesis leads to significant energy savings by avoiding mechanical singularities where actuators would otherwise experience high peak torques.

Results

The framework was validated through several complex 3D paths, including a spatial heart curve and bio-inspired hexapod gaits.

  • Energy Efficiency: Optimized designs achieved a 30% reduction in mechanical energy cost compared to baseline designs that only optimized for path similarity.
  • OveRHex Robot: A hexapod robot designed with this framework achieved a Cost of Transport (COT) of 3.25. This is significantly more efficient than many established small-scale hexapods like RHex (3.74–14) or Weaver (10).
  • Accuracy: Hardware tests using motion capture demonstrated high fidelity, with a normalized Hausdorff distance error of only 6.9% for a target heart trajectory, proving the model's real-world applicability.

Conclusion

This research demonstrates that 1-DoF overconstrained linkages can effectively replicate complex 3D motions. The provided computational tools for collision avoidance and energy optimization make these mechanisms practical for high-performance, efficient robotics. Future work may explore topology optimization to further improve the strength-to-weight ratio of the links.

Summary updated

BibTeX
@misc{pan2025one,
  author = {Jia Pan and Yuping Gu and Bangchao Huang and Haoran Sun and Ronghan Xu and Jiayi Yin and Wei Zhang and Fang Wan and Jia Pan and Chaoyang Song},
  title = {{One-DoF Robotic Design of Overconstrained Limbs with Energy-Efficient, Self-Collision-Free Motion}},
  year = {2025},
  eprint = {2509.22002},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2509.22002},
}
Figure