The VIVID function for numerically continuing periodic orbits arising from grazing bifurcations of hybrid dynamical systems.
I. Ghosh 1 and D.J.W. Simpson 2 1 School of Mathematics and Statistics, University College Dublin, Dublin, D04 V1W8, Ireland 2 School of Mathematical and Computational Sciences, Massey University, Palmerston North, 4410, New Zealand · 2025
Read The VIVID function for numerically continuing periodic orbits arising from grazing bifurcations of hybrid dynamical systems. by I. Ghosh 1 and D.J.W. Simpson 2 1 School of Mathematics and Statistics, University College Dublin, Dublin, D04 V1W8, Ireland 2 School of Mathematical and Computational Sciences, Massey University, Palmerston North, 4410, New Zealand (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-16218.txt.
Introduction
Indranil Ghosh and David J W Simpson fix a blind spot in tracking repeating loops of motion. Many machines mix smooth motion with sudden hits, like a bouncing ball or a chattering tool. Their loops are traced with maps, but tracking fails when a loop just grazes the surface it hits. At that point the math turns singular.
Their fix is a new function named VIVID, short for Velocity Into Variation In Displacement. Instead of hunting fixed points of the map, they follow the zeros of this function, which stay well behaved at a graze. On a model impact oscillator, plain Newton steps then track loops right into the graze. The method also traces curves where loops are born or double, fanning out from one special point. It gives a simple rival to the heavier collocation method now in use.
Catalog record
| CREATOR | I. Ghosh 1 and D.J.W. Simpson 2 1 School of Mathematics and Statistics, University College Dublin, Dublin, D04 V1W8, Ireland 2 School of Mathematical and Computational Sciences, Massey University, Palmerston North, 4410, New Zealand |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-16218
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The VIVID function for numerically continuing periodic orbits arising from grazing bifurcations of hybrid dynamical systems. I. Ghosh 1 and D.J.W. Simpson 2 1 School of Mathematics and Statistics, University College Dublin, Dublin, D04 V1W8, Ireland 2 School of Mathematical and Computational Sciences, Massey University, Palmerston North, 4410, New Zealand Abstract Periodic orbits of systems of ordinary differential equations can be found and continued numerically by following fixed points of Poincaré maps. However, this often fails near grazing bifurcations where a periodic orbit collides tangentially with a boundary of phase space. Failure occurs when the map contains a square-root singularity and the root-finding algorithm searches beyond the domain of viable values. We show that by instead following the zeros of a function that maps Velocity Into Variation In Displacement (VIVID) this issue is circumvented and there is no such failure. We illustrate this with a prototypical one-degree-of-freedom impact oscillator model by applying Newton’s method to the VIVID function to follow periodic orbits collapsing into grazing bifurcations. We also follow curves of saddle-node and period-doubling bifurcations of periodic orbits that issue from a codimension-two resonant grazing bifurcation. The VIVID function provides a simple alternative to the more sophisticated collocation method and enables periodic orbits and their bifurcations to be resolved easily and accurately near grazing bifurcations. 1 Introduction Many mechanical apparatus contain moving parts that bump into one another. Examples include gear assemblies [ 1 , 2 ] , heat exchangers [ 3 , 4 ] , atomic force microscopes [ 5 , 6 ] , church bells [ 7 ] , and the internal elements of self-propelling capsules for endoscopy [ 8 , 9 ] . These examples involve near misses as well as impacts, so are directly relevant to the material in this paper. To predict the motions of such apparatus, one often uses a hybrid model that combines differential equations and maps [ 10 , 11 ] . The differential equations capture the motion between impacts, while the maps encode the effects of the impacts [ 12 , 13 , 14 ] . Each map is applied when the system state reaches the corresponding impacting surface in the phase space of the system. To understand of the typical long-term dynamics, and how this differs for different values of the parameters, we study bifurcations where the dynamics changes in a fundamental way. In particular, grazing bifurcations often correspond to the onset of recurring impacts, and thus are of central importance to many hybrid systems. A stable non-impacting periodic orbit undergoes a grazing bifurcation when it first encounters an impacting surface, necessarily intersecting it tangentially. The bifurcation usually causes the periodic orbit to be destroyed, being replaced by high-period or chaotic motion. Since impacts are highly destabilising, any stable periodic orbits existing near the bifurcation and involving impacts, usually involve only one impact per period. Such orbits are termed maximal [ 15 ] . They involve some p\geq 1 loops close to the path of the non-impacting periodic orbit, where one loop involves an impact and the remaining p-1 loops are near misses. As a maximal periodic orbit emanates from the grazing bifurcation, it is typically unstable but may quickly gain stability through either a saddle-node or period-doubling bifurcation [ 16 , 17 , 18 , 19 ] . To numerically find and continue periodic orbits, both stable and unstable, two methods are commonly used. The collocation method approximates the periodic orbit with an ordered collection of N collocation points (often with N in the order of 100 ) [ 20 ] . These points are adjusted iteratively, e.g. via Newton’s method, until curves through small subsets of consecutive points obey the differential equations to within some tolerance. This method is effective for smooth differential equations and has long been embedded into continuation software including auto [ 21 ] and xppaut [ 22 ] . It can be modified to accommodate hybrid systems by imposing constraints, e.g. that one collocation point belongs to an impacting surface and another collocation point is the image of this point under the map. This is implemented in the software coco [ 23 ] , and we believe that previous works where maximal periodic orbits were continued numerically have exclusively used this approach [ 24 , 25 , 26 , 27 , 28 , 29 ] . The shooting method instead searches for fixed points of an iterate of a Poincaré map. This is effective for smooth differential equations because the induced Poincaré map is smooth (locally), and fixed points are readily located through root-finding. In our setting the required Poincaré map P is piecewise-smooth, having one piece P_{\rm global} corresponding to trajectories that miss the impacting surface, and a second piece P_{\rm global}\circ P_{{\rm disc},R} corresponding to trajectories that experience an impact. A p -loop maximal periodic orbit corresponds to a fixed point of P_{\rm global}^{p}\circ P_{{\rm disc},R} . This composition is smooth, therefore standard root-finding algorithms can, in principle, be employed to compute the fixed point. However, near grazing the root-finding algorithm often attempts to evaluate the composition at points outside its domain of definition, and consequently is unable to locate the fixed point. If the components of P can be smoothly extended through the switching manifold of P , as is the case for grazing-sliding bifurcations of Filippov systems [ 30 ] , this problem can be alleviated by replacing the components of P with their smooth extensions. But for typical grazing bifurcations of impacting hybrid systems, P_{{\rm disc},R} has a square-root singularity and hence does not admit a smooth extension. Essentially the algorithm ‘falls off’ the end of the square-root and terminates unsuccessfully. The purpose of this paper is to overcome this problem by introducing a new function V that is smooth, has no square-root singularity, and whose zeros match one-to-one to the fixed points of P_{\rm global}^{p}\circ P_{{\rm disc},R} . In the context of simple impacting systems, V maps the velocity of the object at impact to the variation in displacement of the object following p loops with one impact. Consequently we call V the VIVID function (abbreviating Velocity Into Variation In Displacement). Below we illustrate the effectiveness of the VIVID function with a harmonically forced linear oscillator with hard impacts. We use parameter values based on the physical experiments described in Pavlovskaia et al. [ 31 ] for which a stable two-loop maximal periodic orbit was observed shortly after the grazing bifurcation. We compute this orbit near grazing and identify saddle-node and period-doubling bifurcations where it gains stability. The corresponding zero of V is readily continued into, and even through, the grazing bifurcation. On the other side of the grazing bifurcation the zero does not correspond to a valid periodic orbit of the hybrid system (here the orbit is virtual [ 32 ] ), but this means that the solution can be continued as close to the grazing bifurcation as desired without loss of accuracy or needing to reduce the step size. This allows us to numerically continue the saddle-node and period-doubling bifurcations, as a second parameter is varied, until these bifurcations terminate at a codimension-two resonant grazing bifurcation. The remainder of this paper is organised as follows. In