DRIVE Through the Unpredictability: From a Protocol Investigating Slip to a Metric Estimating Command Uncertainty
Nicolas Samson 1 {}^{\textbf{1}}, William Larrivée-Hardy 1 {}^{\textbf{1}}, William Dubois 1 {}^{\textbf{1}}, Élie Roy-Brouard 1 {}^{\textbf{1}}, Edith Brotherton 1 {}^{\textbf{1}}, Dominic Baril 1 {}^{\textbf{1}}, Julien Lépine 2 {}^{\textbf{2}}, François Pomerleau 1 {}^{\textbf{1}}, Senior Member · 2025
Read DRIVE Through the Unpredictability: From a Protocol Investigating Slip to a Metric Estimating Command Uncertainty by Nicolas Samson 1 {}^{\textbf{1}}, William Larrivée-Hardy 1 {}^{\textbf{1}}, William Dubois 1 {}^{\textbf{1}}, Élie Roy-Brouard 1 {}^{\textbf{1}}, Edith Brotherton 1 {}^{\textbf{1}}, Dominic Baril 1 {}^{\textbf{1}}, Julien Lépine 2 {}^{\textbf{2}}, François Pomerleau 1 {}^{\textbf{1}}, Senior Member (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-2506-16593.txt.
Introduction
Nicolas Samson and co-authors tackle a hard problem for off-road robots: wheels slip on loose ground, and the slip is hard to predict. Their 2025 paper presents DRIVE, a standard protocol for collecting data that shows how a vehicle responds on each surface. Two ground vehicles, one light and one heavy, were driven on six surfaces from asphalt to ice and mud. The runs covered about five hours and nearly fifteen kilometers.
From this data the team maps how speed commands turn into real motion and steady slip for a skid-steer robot. They also propose a score that estimates how unsure a command is on a given surface. That score can help judge risk before a robot is sent out on rough ground.
Catalog record
| CREATOR | Nicolas Samson 1 {}^{\textbf{1}}, William Larrivée-Hardy 1 {}^{\textbf{1}}, William Dubois 1 {}^{\textbf{1}}, Élie Roy-Brouard 1 {}^{\textbf{1}}, Edith Brotherton 1 {}^{\textbf{1}}, Dominic Baril 1 {}^{\textbf{1}}, Julien Lépine 2 {}^{\textbf{2}}, François Pomerleau 1 {}^{\textbf{1}}, Senior Member |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 93 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-16593
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
DRIVE Through the Unpredictability: From a Protocol Investigating Slip to a Metric Estimating Command Uncertainty Nicolas Samson 1 {}^{\textbf{1}}, William Larrivée-Hardy 1 {}^{\textbf{1}}, William Dubois 1 {}^{\textbf{1}}, Élie Roy-Brouard 1 {}^{\textbf{1}}, Edith Brotherton 1 {}^{\textbf{1}}, Dominic Baril 1 {}^{\textbf{1}}, Julien Lépine 2 {}^{\textbf{2}}, François Pomerleau 1 {}^{\textbf{1}}, Senior Member IEEE Abstract Off-road autonomous navigation is a challenging task as it is mainly dependent on the accuracy of the motion model. Motion model performances are limited by their ability to predict the interaction between the terrain and the UGV , which an onboard sensor can not directly measure. In this work, we propose using the DRIVE ( DRIVE ) protocol to standardize the collection data for system identification and characterization of the slip state space. We validated this protocol by acquiring a dataset with two platforms (from 75 kg 75\text{\,}\mathrm{k}\mathrm{g} to 470 kg 470\text{\,}\mathrm{k}\mathrm{g} ) on six terrains (i.e., asphalt, grass, gravel, ice, mud, sand) for a total of 4.9 h 4.9\text{\,}\mathrm{h} and 14.7 km 14.7\text{\,}\mathrm{km} . Using this data, we evaluate the DRIVE protocol’s ability to explore the velocity command space and identify the reachable velocities for terrain-robot interactions. We investigated the transfer function between the command velocity space and the resulting steady-state slip for a SSMR . An unpredictability metric is proposed to estimate command uncertainty and help assess risk likelihood and severity in deployment. Finally, we share our lessons learned on running system identification on large UGV to help the community. {IEEEkeywords} Command and control, Metrics, Mobile robots, Risk assessment, Robot motion, Terrain factors † † corresponding: Corresponding author: Nicolas Samson (email: [email protected]). † † note: *This research was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) through the BESC M. 1 INTRODUCTION \IEEEPARstart Motion models are essential to multiple key tasks of autonomous navigation such as localization [ 1 ] , path planning [ 2 ] and path following [ 3 ] . For off-road navigation, the accuracy of the motion models is limited by their ability to model the terrain-robot interaction, which is not directly measurable by exteroceptive or proprioceptive sensors [ 4 ] . Poor vehicle-terrain characterization will lead to significant modeling errors, potentially causing system failure [ 5 ] . Furthermore, system identification of off-road UGV is complex due to various factors that modify the traction between the vehicle locomotion mechanism and the terrain [ 6 ] . These factors are called factors of slip for short and can be divided into two groups: internal and external. The internal factors are linked to the driving behavior, the vehicle’s properties and their variation in time. The external factors are linked to the interaction between the UGV and its environment, such as terrain steepness, roughness, and ground hardness. The factors of slip must be well identified when deploying a UGV on a given terrain as they have an important impact on the state variables of the UGV and its 3D navigation dynamics. The identification of these factors of slip requires deploying the UGV in its operational environment and manually driving it for an extended period to gather a dataset that explores this high dimensional system [ 7 ] . This exploration is usually based on two control variables in the body frame: the longitudinal and angular speeds. This identification process remains challenging as the methodology of manual driving is not standardized, which creates a wide variety of motion complexity in the training and validation datasets. The absence of a standard for data gathering strategy, combined with the information limitation of system identification of the terrain- UGV interaction, reduces the reproducibility of the results. Additionally, standardizing this process could help engineers ensure that their systems comply with the ISO 34502:2022(E) standard on autonomous navigation. 2 2 2 “ISO 34502:2022(E): Road vehicles — Test scenarios for automated driving systems — Scenario-based safety evaluation framework”, 2022 To address this issue, [ 8 ] has proposed the protocol called DRIVE ( DRIVE ) , which facilitates and standardizes the terrain-robot system identification as illustrated in Figure 1 . The first step of DRIVE is to identify the true vehicle’s input space, differing from the manufacturer’s specifications. The second step is to determine acceleration and speed limits to prevent damage to the vehicle. The last step is to automatically send commands to the UGV to cover the entire true input space. This differs from the common manual driving approach, which tends to cover only a small proportion of the input space, as shown by the red dots representing our previous work [ 9 ] . Figure 1: Vehicle and terrain characterization done through DRIVE . The manufacturer-defined Naive input-space region is drawn in gray. The vehicle’s true input space, characterized through internal measurements, is shown in orange. Typical human driving is shown in red. The resulting body velocities are represented in green for gravel and blue for snow. This paper is an extension of [ 8 ] ’s work. The DRIVE protocol has been deployed multiple additional times for a total of 13.0\text{\,}\mathrm{km} with a Clearpath Robotic Warthog on five terrains (i.e., sand, ice, asphalt, grass, and gravel) and 1.7\text{\,}\mathrm{km} with a Clearpath Robotic Husky on three different terrains (i.e., grass, asphalt, and mud). This large dataset is the basis of the four contributions of this paper: • the validation that sampling period of DRIVE enables the characterization of the slip state space in steady state. • the analysis of the system identification over six terrains with two robots. • a novel metric related to the unpredictability of autonomous navigation. • lessons learned on running system identifications on large UGV . In this paper, Section 2 presents a review of the dataset-gathering protocol, followed by a review of the factors of slip for UGV and how they have been integrated to different motion models. Section 3 defines the vehicle motion state space and constraints for a typical SSMR , explains the impact of terrain on commands, presents the DRIVE protocol, and presents the formulation of the slip model and the unpredictability metric. Section 4 summarizes the deployments realized and presents the hardware used for the experiments. Section 5 contains the results, which include the analysis of the DRIVE protocol’s ability to explore the steady-state velocities, the transfer functions between the command space and the slip for different terrains, and the unpredictability metric results that can be used to quantify the risk associated with kinetic energy for any robot motion on any terrain. 2 RELATED WORK This section reviews existing dataset-gathering strategies and their limitations. The factors causing vehicle slippage, as identified in the motion modeling literature, are positioned against the general challenge of system identification. Finally, a few proposed metrics to quantify off-road dataset difficulty are presented to show explicitly the need for a metric that quantifies the robot-terrain interactions. 2.A DATASET GATHERING PROTOCOL Although many vehicle motion models require empirical training data, only a few dataset-gathering protocols have been published. [ 10 ] have proposed to maintain a steady forward velocity while slowly increasing angular velocity, enabling the generation of a quasi-steady-state empirical dataset. [ 11 ] have proposed a similar approach with a varying commanded curvature radius to empirically identify the relation between angular velocity