ABSTRACT. Wheel–terrain interaction plays a critical role in predicting the mobility and performance of off-road vehicles operating on deformable terrain. This work presents a soil deformation model designed for real-time simulation, with emphasis on tire-driven soil deformation mechanisms, including bulldozing, compaction, slip sinkage, and multi-pass effects. The proposed approach adopts a 2D height-map representation to update terrain geometry in real time, making it suitable for integration in dynamic simulation environment such as Vortex Studio of CM Labs Simulations, a multi-body simulation in terramechanics framework. The soil mass displaced by the wheel is partitioned and redistributed to soil columns surrounding the wheel footprint in the height-map discretization, using algorithms corresponding to specific soil deformation mechanisms. These algorithms are either theoretically derived or implemented using neural networks trained on DEM-generated data. For example, bulldozing-induced soil deformation is calculated based on a model derived from wedge theory. Slip sinkage-induced soil deformation is estimated using a model developed via symbolic regression. Void ratio changes resulting from soil deformation are also considered, which influence the calculation of soil reaction force. Validation against DEM-generated virtual experiments and experimental data demonstrates good agreement in cases including bulldozing, compaction-induced rut formation, and slip sinkage-induced soil deformation. In summary, the proposed approach can provide computationally efficient and accurate models for the simulation of off-road vehicle mobility, heavy equipment, and construction machinery.

ABSTRACT. This paper presents an experimental and numerical investigation of regolith dust mobilized by rover wheel locomotion on the lunar surface, using the lunar regolith simulant LHS-1. Regolith dust lifted during rover traversal is a critical concern because it can penetrate or adhere to mechanical and thermal components, potentially causing performance degradation or failures, and may also affect nearby surface assets and astronauts performing extravehicular activities. While numerous terramechanics-based studies have investigated wheel-soil interaction in terms of mobility performance metrics such as drawbar pull and sinkage, relatively few have focused on the subsequent behavior of regolith dust mobilized by wheel motion. To address this gap, a ground-based single-wheel test rig operable under vacuum conditions was developed, and the dynamics of mobilized regolith dust were experimentally observed using a high-sensitivity, high-speed imaging system. To enable detailed observation at arbitrary slip ratios, the regolith simulant bed was translated to reproduce relative motion equivalent to vehicle forward speed. Discrete element method (DEM) simulations were conducted and qualitatively compared with experimental observations. In addition, numerical simulations under one-sixth gravitational acceleration were performed to approximate lunar gravity, and a method was proposed to estimate the spatial extent of regolith dust mobilization caused by wheel traversal. The results demonstrate that the proposed experimental–numerical framework enables quantitative evaluation of regolith dust behavior associated with wheel–soil interaction in lunar environments.

ABSTRACT. In any vehicle dynamics simulation model, the tyre model is of upmost importance as it is the only component in contact with the terrain. In many cases engineers under estimate the contribution the tyre model makes to the reliability and accuracy of the simulation results. On larger off-road vehicles this is the case due to the limited tyre data and models available to the engineers as parameterizing tyres are not trivial and can be very expensive. On very large tyres the infrastructure required to conduct tyre parameterization tests are not available. Vehicles used on soft terrain are also often used on hard terrain. Tyre parameters are measured on hard terrain as the tyre carcass construction, inflation pressure and vertical load on the tyre dictates the linear section of the longitudinal and lateral tyre parameters. The maximum longitudinal and lateral load generated by the tyre is dictated by the surface roughness of the terrain over which the tyre transverse. In this study we conduct in-situ tests on a vehicle on a specific surface. The results are compared to laboratory tests conducted on the same tyre and same surface. This method enables engineers to obtain first order tyre parameters to validate simulation models with limited and non-specialized equipment. The aim of this study is to validate the test methodology of the implementation on large off-road vehicles with tyres too large to tests in current laboratories.

ABSTRACT. The Dakar Rally is a race that takes place on unpaved terrain, and it is said that about 40% of the course consists of sandy terrain. Because desert sand is dry and very loose, tires tend to slip on sandy terrain, which can cause them to sink into the sand. This study investigated tread patterns that provide sufficient friction on sandy surfaces while reducing tire sinkage. Many existing tire tread patterns are designed to push loose surface soil to the sides and grip relatively firm ground to generate traction. However, on soft sand, this design can cause sand to be pushed out from under the tire tread, potentially causing the tires to sink further. For this reason, we proposed several tread patterns designed to ensure sufficient friction while preventing outward sand movement and conducted experiments to verify their effectiveness. In the single-wheel running test, miniature tires were used. The tires are directly driven by a motor with a reduction gear on “Toyoura sand.” The tread pattern was designed not to move sand outward as the tire rotates, but to trap it inward, with the aim of improving traction by the pressure effect of the sand and preventing tire sinkage. The results showed that the inverted U-shaped pattern demonstrated relatively good traveling performance at slip ratios of 40% to 50%. However, when the wheel rotation speed was high and the slip ratio exceeded 55%, the U-shaped pattern proved effective. The accumulation of sand in the center caused the pattern to become clogged, preventing sufficient traction from being generated. This study demonstrated that adopting a block-shaped arrangement—which allows sand to flow appropriately while still gathering it in the center of the tire—is effective.

ABSTRACT. A track drawbar pull model that considers the effects of grouser shape is essential for accurately calculating the driving performance of tracked vehicles on sandy terrain. The classical model calculates the drawbar pull as the integral of the shear stresses generated at the track-terrain interaction patch. However, its accuracy and applicability are limited because the track shape in the model is simplified as a flat plate and the effects of the grousers are not considered. In this study, we aim to extend the shear stress model proposed by Janosi et al. to consider the effect of grouser shapes. First, we examine the physical meanings of the model parameters, namely soil deformation modulus and internal friction angle, and then redefine these parameters in accordance with the grouser shapes. A new shear test device called “grouser-shoe shear test” is elaborated to identify these parameters. The device consists of multiple actual track shoes with grousers connected together, and is controlled to move horizontally with a certain amount of vertical load applied. The parameters being equivalent to the shear deformation modulus and friction angle are identified from the relationship between the vertical load and horizontal shear stress. The calculated track drawbar pull coefficient using these identified values closely matched with the shape of the graph obtained from drawbar pull experiments using a track unit with rectangle grouser shoes. These results suggest that proposed parameters identification approach enables prediction of the driving performance of grousered-track units with reasonable accuracy.

ABSTRACT. The growing demands for efficiency, safety and comfort in modern agriculture make it essential to accurately predict the handling behaviour of agricultural machinery. A key element of vehicle development is the modelling of tyre behaviour, for which the Institute of Agricultural Engineering at the University of Hohenheim, Stuttgart, Germany, developed the Hohenheim Tyre Model (HTM). The model enables comprehensive vehicle dynamics analysis through co-simulation with multi-body systems and can be parameterised using the institute's own test benches. A persistent challenge in tyre modelling is the limited availability of test benches for tyres with large diameters, as modern agricultural tyres increasingly exceed the dimensional and load limits of existing test rigs. To address this limitation, this paper presents a method for determining HTM parameters directly from road tests using an instrumented tractor equipped with wheel force transducers. Lateral force-slip angle characteristics, longitudinal force-slip characteristics, vertical tyre parameters, and radial runout were determined for the front and rear tyres and compared with test bench measurements for the front tyre. The results show that the road test method offers a viable alternative to test bench measurements for most parameters, although the precision is limited in certain ranges, particularly for vertical parameters under active suspension conditions. The method thus provides a practical means of parameterising the HTM for large modern tractor tyres that cannot be accommodated by existing test benches. Future work will extend the HTM to include the simulation of soft soils, representing the interaction between tyres and the terrain under variable field conditions.

ABSTRACT. Vehicle–terrain simulations rely on tyre models to supply the forces and moments that drive vehicle dynamics, making accurate local contact patch information essential for tyre–terrain modelling. This paper presents a framework for reconstructing the two- dimensional normal pressure field in a static tyre contact patch from full-field inner-surface displacement measurements to compare and verify analytical pressure fields with measurement- informed tyre contact pressure in brush tyre models. The proposed approach uses Tyre-Terrain Camera (T2Cam) measurements to define the measured radial displacement, measurement-informed boundary conditions and the estimated contact patch. The pressure field is represented by low-order longitudinal basis functions on coupled lateral strips. A calibrated beam-strip forward model maps the pressure coefficients to predicted inner-surface displacement, after which the inverse reconstruction is posed as a constrained, regularised quadratic programme that enforces non-negativity, confines pressure to the contact patch area derived from T2Cam, and matches the known vertical load. The planned experimental tests use a Static Tyre Test Rig with T2Cam, Tekscan pressure measurements, and a wheel force transducer to calibrate the forward operator and validate the reconstructed pressure field before the method is used without direct pressure measurements. A preliminary MATLAB transient dynamic brush model has been implemented with bristle recycling and a dynamic friction state formulation. These preliminary results demonstrate the downstream modelling interface that will later replace analytical pressure fields with reconstructed pressure fields from the inverse solver.

ABSTRACT. Accurate localization of forest vehicles is challenging because GNSS signals are often degraded by canopy cover and terrain, while wheel slip and dynamic disturbances reduce the reliability of dead-reckoning sensors. An asynchronous Extended Kalman Filter (EKF) is proposed for estimating the 2D state of a center-articulated forest vehicle using GNSS RTK measurements at 5 Hz, an IMU at 360 Hz, wheel speed, and articulation angle measurements at 50 Hz. The filter estimates the vehicle position and heading while accounting for the nonlinear kinematics of articulated motion and sensor uncertainty. Field data were collected using the Arctic Off-Road Laboratory on an asphalt parking lot under open sky conditions and replayed in simulation, where synthetic GNSS outages of 60 s and 120 s were introduced to evaluate dead-reckoning performance. During 60 s outages, the autonomous rectangular trajectory traveled 37.8 m and achieved a mean positional error of 0.22 m, a maximum positional error of 1.91 m, and heading errors below 1°, while the manually driven rectangular trajectory traveled 65.7 m with mean positional errors of 0.31–0.50 m and a maximum positional error of 2.01 m. The wavy trajectory traveled 63.7 m and produced the largest drift, with mean positional errors ranging from 0.39–1.23 m, depending on the IMU configuration, and maximum positional errors of 4.20–7.38 m. IMU integration generally improved performance during longer outages. Overall, the proposed EKF improves robustness to GNSS outages and provides a foundation for future localization methods for articulated forest vehicles. Future work will focus on filter parameter optimization and experimental validation in real off-road environments.

ABSTRACT. Quadruped robots are expected to be useful in rough terrain where it is difficult for people or vehicles to access, such as disaster sites. While they excel at moving over hard, uneven ground, their spherical feet—which have a small contact area—sink into the ground when walking on soft terrain. However, on soft slopes, driving the spherical feet into the ground like piles can suppress slippage more effectively than feet with a larger contact area. Therefore, changes in walking performance are expected depending on the foot trajectory. In this study, we investigated the foot trajectories of spherical feet that improve walking performance on soft ground by analyzing bearing capacity and propulsive force from single-leg tests, as well as walking distance from a quadruped robot. Initially, we measured the forces acting on the leg on sandy ground using a single-leg testing apparatus modeled after the legs of a quadrupedal robot. We measured the effects on bearing force and propulsive force by altering the foot trajectory during penetration and push-off. The results revealed that, for the same penetration depth, both forces increased as the penetration angle decreased. Finally, we verified the effects of the trajectory by applying the foot trajectory from the single-leg test to a quadruped robot and measuring the walking distance. On an inclined surface, walking distance improved due to an increase in propulsive force resulting from diagonal foot penetration and a foot trajectory that positioned the center of gravity near the midpoint between the front and rear foot contact points.

ABSTRACT. Autonomous slip detection allows a rover to detect potentially dangerous terrain and take precautions. While visual odometry slip estimation solutions exist, they require significant computational resources and may not be available on upcoming micro-rover missions. Previous studies have shown promise in the use of Machine Learning algorithms to process Inertial Measurement Unit (IMU)-measured vibration data to detect and classify slip events. This research develops a low-latency and computationally efficient vibration-based system to detect wheel-terrain slip events for skid-steer rovers with modest hardware requirements. To this end, vibration datasets corresponding to various wheel-terrain slip values are generated. A Husky rover with IMUs is used in indoor and outdoor test environments where slip is induced at specific values by mechanically constraining the rover. The vibration datasets are used to train and validate a Support Vector Machine classifier to differentiate abnormally high slip events from normal low slip. The training is done with various sensor channels, sampling frequencies, and sample window durations. The performance of the system is then evaluated in order to find which combinations of parameters are effective and to identify trade-offs in performance. Field tests are performed on novel terrain not seen in training, and the parameter trade-off analysis holds up robustly. For example, a 100Hz IMU with a 0.25s sampling window achieves excellent performance in the field (> 95% area under the Receiver Operating Characteristic (ROC) curve), better than a 10Hz IMU with a 5s sampling window.

ABSTRACT. The study addresses the prediction of terrain-induced mobility conditions for optionally manned ground vehicles operating in autonomous mode using proprioceptive sensing. The objective is to estimate mobility degradation based solely on internal vehicle signals to support control adaptation in unstructured environments. Experimental data were acquired using the TAERO platform with varying payload configurations during field trials conducted on compact soil, loose surfaces, and vegetation-covered terrain. The sensing configuration included wheel encoders and an inertial measurement unit with integrated GNSS, providing time-series data describing vehicle–terrain interaction. Signal features such as slip ratio, acceleration variance, and frequency-domain characteristics were extracted and correlated with observed mobility conditions. The results indicate that distinct terrain-induced mobility states can be identified using proprioceptive measurements, with clear separation between low-resistance and high-resistance conditions. Increased slip ratios and vibration energy were consistently associated with reduced mobility and approaching stability limits. The proposed approach enables early detection of mobility degradation prior to critical loss of traction or stability, which is important for control system behaviour in autonomous mode. The results provide a basis for integrating mobility estimation into terrain-aware control strategies and mobility-based performance evaluation of optionally manned UGVs

ABSTRACT. In almost all engineering endeavours the use of computational modelling is crucial in ensuring cost effective and efficient solutions. Autonomous Ground Vehicle (AGV) systems are no different and is some ways even more beneficial as it allows the autonomy stack capabilities to be evaluated, updated and then reevaluated digitally and updated before being implemented on the physical platform. This paper presents the initial effort by the CSIR to develop a robust, efficient and effective AGV capability building methodology starting with freeware offering the capability to model AGV in digital twin operational environment. Using simple ideal perception sensors and limited physics-based vehicle models in Gazebo to develop a lead vehicle-follower capability for the existing CSIR Robot Operating Systems (ROS) 2 autonomy stack for off-road applications on a newly built four wheeled electric drive platform. A digital twin environment was used providing both built environment with concrete buildings, concrete and tar road surfaces but also undulating uneven wooded grasslands to test the system capabilities digitally. This digital twin also offered the opportunity to verify and validate the limited AGV model both in on and off-road environments. The training of the stack used a combination of real-world photographs of the lead vehicle and a detailed scanned model. The initial results confirmed that many development issues can be identified and corrected in the model before development. The use of simple vehicle models requires extensive tuning of the system control parameters on the physical systems with regular functionality trials before deployment; however, this is generally required irrespective to verify both the system safety as well as functionality.