Lie Group Variational Collision Integrators for a Class of Hybrid Systems

Khoa D. Tran, Melvin LeokSIAM J. Appl. Dyn. Syst 2023

TL;DR

This paper introduces a novel Lie group variational collision integrator (LGVCI) for simulating 3D rigid-body collisions, particularly those involving sharp corners. The LGVCI offers symplectic and momentum-preserving properties, ensuring accurate long-term energy conservation, and effectively handles complex impact scenarios without resorting to nonsmooth analysis.

What Problem Does This Paper Address?

Simulating the dynamics of rigid bodies undergoing collisions, especially those with sharp corners, presents a significant challenge in physics engines and robotics. Traditional methods often struggle with the discontinuous nature of collisions and the complexities introduced by non-smooth geometries. Existing approaches can be computationally expensive, lack global descriptions of motion, or fail to preserve crucial physical properties like energy and momentum over long simulation times. This paper aims to bridge this gap by developing a robust and accurate numerical integrator that can handle these complex collision scenarios on Lie groups, providing a more faithful representation of real-world physics.

Methods

The authors extend the framework of variational integrators and Lie group integrators to develop their Lie group variational collision integrator (LGVCI). Key techniques include:

  • Nonsmooth Lagrangian Mechanics: This forms the theoretical foundation for deriving equations of motion and jump conditions during collisions.
  • Lie Group Integrators: By leveraging the geometry of the Special Euclidean group SE(3), the integrator provides a global description of the rigid-body's position and orientation, avoiding issues with local representations like Euler angles.
  • Variational Approach: The integrator is derived from a discrete variational principle, which inherently leads to symplectic and momentum-preserving properties.
  • e-rounding Regularization: To handle sharp corners, the authors introduce an "e-rounding" technique. This smooths the body's boundary by a small parameter $\epsilon$, effectively regularizing the signed distance function (SDF) and avoiding the need for complex nonsmooth convex analysis at impact points.
  • Collision Detection Function: A signed distance function is used to detect collisions and determine impact times and configurations.

Datasets

Not discussed in the paper. The paper focuses on developing and testing a numerical method rather than training on a dataset.

Key Insights and Ideas

  • Global Description via Lie Groups: Using SE(3) for configuration space provides a more elegant and robust way to describe rigid-body motion, especially during complex maneuvers and collisions, compared to local representations.
  • Variational Integrators for Collision Handling: The variational framework naturally enforces conservation laws (symplectic, momentum-preserving), which are crucial for accurate long-term simulations of systems with many collisions.
  • e-rounding as a General Solution for Sharp Corners: This regularization technique is a clever way to circumvent the difficulties of nonsmooth analysis for corner impacts, offering a unified approach for various convex shapes. It allows the use of standard variational integrator machinery even in the presence of sharp geometric features.
  • SDF as a Unified Tool: The signed distance function proves to be a versatile tool, enabling collision detection, determining impact points, and facilitating the regularization of complex geometries.

Key Results

The paper presents a novel LGVCI that demonstrates:

  • Symplectic and Momentum-Preserving Properties: Verified through numerical experiments, these properties are essential for long-term simulation stability.
  • Accurate Energy Conservation: The integrator exhibits excellent near-energy conservation, with only a slight drift observed over long simulation times due to numerous collisions and the discrete nature of the time-stepping.
  • Effective Handling of Sharp Corners: Numerical experiments with a cube and other polyhedra show that the e-rounding approach successfully resolves collisions at sharp corners without introducing numerical instabilities.
  • Sensitivity to Initial Conditions: The simulations confirm that the integrator accurately captures the sensitive nature of these hybrid systems to initial conditions, as demonstrated by the divergence of trajectories after a few collisions.

Limitations and Future Work

  • Zeno Phenomenon: While the current implementation with smooth or e-rounded bodies avoids the Zeno phenomenon (infinite collisions in finite time), the authors note that it could arise with non-smooth rigid bodies. They suggest remedies like breaking loops or returning outputs when a threshold is exceeded.
  • Elastic Collisions: The current work focuses on elastic collisions. Future work could extend the LGVCI to handle inelastic collisions and friction.
  • External and Contact Forces: Incorporating external forces and more general contact scenarios is a natural extension.
  • Elastic Bodies: Formalizing and integrating elastic bodies (e.g., hyperelastic materials) into the framework is a promising direction.
  • Convex-Nonconvex Rigid-Bodies: Extending the analysis to collisions between convex and nonconvex rigid-bodies, and systems exhibiting energy dissipation, is a key area for future research.
  • Geometric Control and Optimal Control: The authors highlight interesting applications in geometric control and optimal control on SE(3) with unilateral constraints.

Summary updated

Abstract

The problem of 3-dimensional, convex rigid-body collision over a plane is fully investigated; this includes bodies with sharp corners that is resolved without the need for nonsmooth convex analysis of tangent and normal cones. In particular, using nonsmooth Lagrangian mechanics, the equations of motion and jump equations are derived, which are largely dependent on the collision detection function. Following the variational approach, a Lie group variational collision integrator (LGVCI) is systematically derived that is symplectic, momentum-preserving, and has excellent long-time, near energy conservation. Furthermore, systems with corner impacts are resolved adeptly using $ε$-rounding on the sign distance function (SDF) of the body. Extensive numerical experiments are conducted to demonstrate the conservation properties of the LGVCI.
BibTeX
@article{tran2023lie,
  author = {Khoa D. Tran and Melvin Leok},
  title = {{Lie Group Variational Collision Integrators for a Class of Hybrid Systems}},
  journal = {SIAM J. Appl. Dyn. Syst},
  year = {2023},
}
Figure