🔢 mathematics

Analysis and experiments of the dissipative Twistcar: direction reversal and asymptotic approximations

This paper presents a theoretical and experimental study of a dissipative Twistcar model that incorporates rolling resistance, deriving asymptotic approximations for its steady-state dynamics to predict and experimentally demonstrate the reversal of motion direction by adjusting the vehicle's geometric and mass properties.

Original authors: Rom Levy, Ari Dantus, Zitao Yu, Yizhar Or

Published 2026-08-11

📖 3 min read🧠 Deep dive

CC BY 4.0

Original authors: Rom Levy, Ari Dantus, Zitao Yu, Yizhar Or

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a world where robots don't just follow a pre-programmed map but can wiggle, twist, and scoot around using only their own body movements, much like a snake slithering through grass or a fish darting through water. This is the realm of "underactuated" systems, where a machine has fewer motors than it has ways to move. It's a bit like trying to drive a car that has no steering wheel; instead, you have to twist the chassis back and forth to make it turn. These systems are governed by "nonholonomic" rules, a fancy way of saying that while the robot can move in many directions, it can't just slide sideways like a crab; it has to roll forward or backward to get anywhere. Scientists care deeply about these wiggly machines because they are the key to building agile robots that can navigate messy, real-world environments where traditional wheels get stuck. But there's a catch: in the perfect, frictionless world of computer theory, these wiggling robots often behave strangely, accelerating forever until they break the laws of physics. The big question is: how do we make these theoretical wiggles match the messy, sticky reality of the real world?

This paper dives into that exact puzzle using a classic toy called the "Twistcar," which looks like a two-part scooter with a steering joint in the middle. The researchers wanted to figure out why, in real life, this toy doesn't spin out of control, and whether they could make it change its mind and drive backward just by shifting its weight. They started by building a mathematical model that included "rolling resistance"—the idea that wheels get a little tired and slow down due to friction, rather than rolling forever on ice. When they ran simulations with this friction, the toy behaved nicely, settling into a steady, rhythmic wobble instead of flying apart.

The most exciting discovery came when they played with the toy's mass. The team built a real, modular robot version of the Twistcar that could have a heavy block added or removed from its body. They found that by simply moving this weight, they could flip the robot's direction. Without the extra weight, the robot would scoot backward; with the weight added, it would suddenly start moving forward. It's as if the robot realized, "Oh, I'm heavier now, I should go the other way!" They proved this phenomenon both in their computer models and with their physical robot, showing that the direction of motion depends heavily on how the mass is distributed.

However, the road to this discovery wasn't perfectly smooth. The researchers found that their simple friction model didn't perfectly match the real robot's behavior. When they looked closer, they realized two things were messing with the results: the wheels were slipping a tiny bit sideways (skidding), and the floor they were testing on wasn't perfectly flat. Even a tiny slope of just 1 degree was enough to change the robot's speed significantly. By tweaking their math to account for these "imperfections," they got much closer to the real-world results. Ultimately, the paper suggests that while we can predict these wiggly robots quite well, we have to be very careful about the tiny details of friction and the ground they roll on to get the direction exactly right.

Technical Summary: Analysis and Experiments of the Dissipative Twistcar

Problem Statement
The Twistcar is a classical underactuated, two-link wheeled vehicle inspired by a riding toy, modeled as a nonholonomic system with periodic actuation of the relative angle between its links. Previous theoretical analyses of the Twistcar and similar systems (e.g., the Snakeboard) often relied on non-dissipative models. These models, which assume ideal nonholonomic no-skid constraints, predict unbounded, diverging oscillatory velocities and orientation angles. However, experimental observations contradict this, showing bounded steady-state motion. The discrepancy arises because ideal nonholonomic constraints do not perform mechanical work, failing to account for energy dissipation mechanisms present in real-world systems, such as rolling resistance and wheel skidding. This work addresses the need to incorporate energy dissipation into the theoretical model to accurately predict bounded dynamics and phenomena such as direction reversal.

Methodology
The authors employ a combined approach of theoretical analysis, numerical simulation, and experimental validation:

Dynamic Formulation: A planar two-link model is formulated using the Euler-Lagrange method. The system incorporates nonholonomic no-skid constraints in the lateral direction and introduces viscous dissipation to account for rolling resistance in the direction of wheel motion. The system is driven by a time-periodic sinusoidal input of the steering angle.
Asymptotic Analysis: To obtain explicit analytical expressions, the authors perform a perturbation expansion for small-amplitude actuation. The dynamics are reduced by enforcing constraints, leading to a simplified system of equations. The analysis derives leading-order asymptotic expressions for the vehicle's forward speed and orientation angle.
Numerical Simulation: The full dynamic equations are solved numerically using MATLAB's ode45 solver. Simulations are conducted for both dissipative and non-dissipative models to compare steady-state behaviors and investigate the conditions for direction reversal by varying geometric and mass parameters.
Experimental Validation: A modular robotic prototype was constructed using a goBILDA chassis, a servo motor, and an Arduino controller. The robot features a reconfigurable structure allowing the addition or removal of a massive block to shift the center of mass. Motion tracking was performed using a VICON system. The authors utilized parameter fitting to estimate the viscous dissipation coefficient by comparing experimental data with numerical simulations.

Key Contributions and Results

Bounded Dynamics with Dissipation: The study demonstrates that incorporating viscous rolling resistance into the nonholonomic model resolves the unbounded behavior predicted by ideal models. The dissipative model yields bounded, steady-state oscillatory motion that aligns with physical reality.
Direction Reversal Mechanism: Theoretical analysis and simulations reveal that the direction of the vehicle's net motion can be reversed by altering its geometric and mass properties. Specifically, the sign of the mean steady-state velocity depends on a derived condition ( $\xi$ ) involving mass distribution, link lengths, and moments of inertia.
Experimental Demonstration: The robotic prototype successfully validated the direction reversal phenomenon. By shifting the center of mass (adding a mass block to one link), the robot's motion was observed to switch from backward (negative velocity) to forward (positive velocity), confirming the theoretical predictions.
Parameter Fitting and Model Limitations: The authors estimated the viscous dissipation coefficient through parameter fitting. While the model captured the general behavior, discrepancies remained, particularly regarding the orientation angle drift. The study identified two significant unmodeled effects:
- Wheel Skidding: Measured skid velocities were small but non-negligible. Incorporating skid dissipation into the model improved agreement with experimental speed data.
- Ground Levelness: Simulations showed that even minor slopes (as small as $1^\circ$ ) could significantly alter the robot's speed and potentially reverse its direction. This sensitivity explains the variance observed in experiments conducted under identical input parameters.
Asymmetric Input Analysis: The paper extends the asymptotic analysis to asymmetric inputs (non-zero mean steering angle), deriving expressions for the resulting curved trajectory and mean curvature of the path.

Significance
The paper establishes that energy dissipation is a critical factor in the dynamics of underactuated nonholonomic vehicles like the Twistcar. By integrating viscous rolling resistance while preserving nonholonomic constraints, the authors provide a more accurate theoretical framework that predicts bounded steady-state behavior. The work successfully bridges the gap between theoretical models and experimental reality, demonstrating that direction reversal is a robust phenomenon achievable through geometric and mass manipulation. The findings highlight the sensitivity of such systems to physical imperfections, including wheel skidding and terrain irregularities, suggesting that future control strategies and model refinements must account for these factors to achieve precise motion planning and stabilization.

More like this