Online Slip Detection and Friction Coefficient Estimation for Autonomous Racing
Christopher Oeltjen, Carson Sobolewski, Saleh Faghfoorian, Lorant Domokos, Giancarlo Vidal, Ivan Ruchkin · 2025
Read Online Slip Detection and Friction Coefficient Estimation for Autonomous Racing by Christopher Oeltjen, Carson Sobolewski, Saleh Faghfoorian, Lorant Domokos, Giancarlo Vidal, Ivan Ruchkin (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-2509-15423.txt.
Introduction
Christopher Oeltjen leads a team that watches for tire slip in self driving race cars. Grip between tire and road sets how fast a car can safely go. That grip cannot be sensed directly with standard gear. Older ways to gauge it need detailed car models or large training data. The team wanted a light method that works live.
Their method uses only motion and laser sensors plus the car's own commands. It spots slip by comparing the motion ordered with the motion seen. When there is no slip, it reads grip straight from the car's speedup. Tests with a small race car at varied grip levels matched ground truth well. The method is cheap to run and easy to deploy.
Catalog record
| CREATOR | Christopher Oeltjen, Carson Sobolewski, Saleh Faghfoorian, Lorant Domokos, Giancarlo Vidal, Ivan Ruchkin |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-15423
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Online Slip Detection and Friction Coefficient Estimation for Autonomous Racing Christopher Oeltjen, Carson Sobolewski, Saleh Faghfoorian, Lorant Domokos, Giancarlo Vidal, Ivan Ruchkin Abstract Accurate knowledge of the tire–road friction coefficient (TRFC) is essential for vehicle safety, stability, and performance, especially in autonomous racing, where vehicles often operate at the friction limit. However, TRFC cannot be directly measured with standard sensors, and existing estimation methods either depend on vehicle/tire models with uncertain parameters or require large training datasets. In this paper, we present a lightweight approach for online slip detection and TRFC estimation. Our approach relies solely on IMU and LiDAR measurements and the control actions, without special dynamical or tire models, parameter identification, or training data. Slip events are detected in real time by comparing commanded and measured motions, and the TRFC is then estimated directly from observed accelerations under no-slip conditions. Experiments with a 1:10-scale autonomous racing car across different friction levels demonstrate that the proposed approach achieves accurate and consistent slip detections and friction coefficients, with results closely matching ground-truth measurements. These findings highlight the potential of our simple, deployable, and computationally efficient approach for real-time slip monitoring and friction coefficient estimation in autonomous driving. Keywords: Slip detection, friction coefficient estimation, autonomous racing I INTRODUCTION The effectiveness of vehicles depends on accurate knowledge of key vehicle and environment parameters, among which the tire-road friction coefficient (TRFC) is one of the most critical [ 1 , 2 ] . The TRFC defines the upper limit of forces that can be transmitted between tires and road, thereby constraining braking, steering, and overall vehicle stability [ 3 ] . Accurate and real-time estimation of this parameter is essential not only for human drivers, but increasingly for automated decision-making, and trajectory planning in autonomous vehicles [ 4 ] . In the domain of autonomous racing, operating at or near these friction limits is essential to achieve competitive lap times. Online slip detection is therefore a crucial capability: it provides feedback for exploiting available traction while maintaining stability, and it supports higher-level functions such as adaptive control, trajectory planning, and race strategy. Moreover, since tire conditions strongly influence traction availability, slip detection also plays a key role in monitoring and managing the effective use of tires during high-performance driving [ 5 ] . Unlike velocity or acceleration, however, the TRFC cannot be measured directly by standard onboard sensors. Specialized sensors capable of direct measurement exist; however, their high cost and integration challenges prevent their widespread use. This limitation has motivated extensive research into indirect estimation methods [ 6 ] . Broadly, TRFC estimation approaches can be categorized into cause-based and effect-based approaches. Cause-based approaches aim to characterize road surface properties directly, using cameras, laser scanners, or acoustic sensing [ 7 , 8 ] . Although these approaches can provide useful surface classification, they typically yield only coarse friction estimates and often require large training datasets. Effect-based approaches instead infer TRFC from the dynamic responses of the vehicle, exploiting measurable quantities such as wheel slip, sideslip angle, yaw rate, or acceleration [ 9 , 10 ] . These approaches tend to be more accurate, as they incorporate the actual interaction between tire and road, but they tend to rely heavily on vehicle or tire models whose parameters may vary with conditions. Within effect-based approaches, a large body of research has adopted model-based techniques such as Kalman filtering. Extended Kalman Filters (EKF) [ 9 ] , Unscented Kalman Filters (UKF) [ 11 ] , and Cubature Kalman Filters (CKF) [ 12 ] have all been applied to TRFC estimation, fusing standard vehicle sensors with simplified models. These approaches can significantly improve estimation accuracy, but they remain model-dependent and are sensitive to parameter uncertainty and nonlinearities. More recently, model-free approaches have been explored as alternatives. These approaches aim to bypass parametric modeling by directly mapping sensor signals to friction estimates, and have also been proposed more broadly for vehicle dynamics modeling, such as physics-constrained neural networks for autonomous racing [ 13 ] , highlighting the growing interest in model-free and data-driven approaches. Examples include learning-based estimators using neural networks trained on vehicle accelerations, data-driven terrain classification methods, and vision–language model-based clustering for friction estimation in off-road contexts [ 14 ] . Such approaches show promise in handling complex, nonlinear interactions without explicit models, but they generally require extensive datasets and may lack generalization to unseen conditions. In this paper, we propose a lightweight approach for slip detection and TRFC estimation that does not require tire or dynamical models, parameter identification, or large training datasets. Instead, it leverages only standard onboard motion sensors (IMU and LiDAR), making it simple and practical to deploy. We first design a slip detection module, which determines the precise moment when the tires lose grip. At that moment, our friction estimation module computes the maximum friction coefficient directly from the observed vehicle motion. Despite its lightweight nature, our approach detects the slip of a 1:10-scale autonomous racing car on a closed-loop track with high accuracy and minimal delay. Furthermore, our friction estimation method accurately estimates the tire-road friction coefficient, effectively differentiating between two test surfaces of varying friction. The contributions of this paper are: • A lightweight slip detection technique using only IMU and LiDAR readings and control actions. • A real-time estimation method for TRFC without explicit tire/vehicle modeling or training data. • Experimental validation demonstrating reasonable online performance of our approach. The rest of this paper is organized as follows. Section II provides the key definitions and formulates the slip detection and friction estimation problems. Section III reviews the related work. Section IV presents the proposed lightweight approach, and Section V reports its experimental validation. Finally, Section VI concludes the paper. II PROBLEM FORMULATION This section formalizes the key concepts and notation underlying our approach. We interleave general definitions and the specifics of autonomous racing, following standard notions in vehicle dynamics by Gillespie [ 15 ] and Rajamani [ 16 ] . II-A Vehicle Level Definition 1 (Plant) The plant P is the physical system under consideration, including its geometry and physical parameters. The plant has a center of mass (CoM), where the total mass m is assumed to be concentrated. A right-handed body-fixed 3D coordinate system B=\{X,Y,Z\} is attached at point CoM , as shown in Fig. 1 : • Longitudinal axis X points forward along the chassis • Lateral axis Y points leftward at 90° to axis X • Normal axis Z points upward In our case, the plant is a 1/10-scale autonomous racing car (also referred to as the vehicle), and it is characterized by a set of configuration , inertial , and geometric properties. The configuration is defined by four tires in contact with the road. Each tire has an effective tire radius r_{e} , defined as the distance from the tire center to the contact point as shown in Fig. 2 . The vehicle is rear-wheel drive (RWD), while the front axle is actuated (steerable). The inertial properties of the vehicle include its total mass m and the yaw moment of inertia I_{z} about the normal axis Z . The geometric properties of our vehicle, illustrated in Fig. 1 , consist of the distances from