The automotive industry is heading towards a more objective approach to vehicle testing, but subjective evaluation is still an important part of the development process. Subjective evaluation in physical testing has environmental implications and is dependent on ambient conditions. A more repeatable, faster, safer and more cost-effective tool for subjective evaluation is to use moving base driving simulators. The motion cueing algorithms (MCA) maps the movement of the vehicle into the limited space of the simulator. The choice of reference point, that is, where on the vehicle to sample the motion to feed to the MCA and the alignment of the axis of rotation of the simulator cabin is still an open topic. This paper investigates the choice of reference point and corresponding simulator longitudinal axis of rotation in roll using two methods. The first method uses a linearised model of the combined system of vehicle, simulator and vestibular models. The second method, to position the cabin longitudinal axis of rotation, is based on offline optimisation. The linear model can capture important characteristics of the specific forces and rotations that are fed to the driver through the motion cueing algorithms and offers a method to objectively analyse and potentially tune the motion cueing. The analysis is further complemented with a subjective evaluation of corresponding settings. The results from the linear model, the offline optimisation and the subjective evaluation shows that a reference point at the driver’s head has a clear advantage over the full frequency range compared to a reference point in the chassis roll axis and that the positioning of the cabin longitudinal axis of rotation has a significant effect on the perceived vehicle characteristics.

This paper investigates the tuning of motion cues in dynamic driving simulators for EPAS (Electronic power assisted steering) tuning in winter conditions. The study investigates the differences in frequency content of the vehicle states yaw rate and lateral acceleration between dry and winter EPAS tuning. Based on the results from this investigation, which shows an increased spectral density of low frequency content in both yaw rate and lateral acceleration, coordinated tilt is added to the motion cueing, to give the driver low frequency lateral acceleration feedback. The tilt coordination filter is tuned using offline optimisation based on logged data. The resulting MCA is evaluated objectively using a linear model of a driving simulator and subjectively through driving around a winter test track with six test drivers. The test is conducted using a pairwise comparison of two different settings, one setting without and another with added tilt coordination. Objective metrics shows reduction in lateral false cues, increased correlation between actual vehicle acceleration and simulator acceleration and an increased spectral density below 0.30 Hz. The pairwise comparison and the commentary feedback shows potential in adding tilt coordination with more drivers favouring tilt coordination, however statistical significance cannot be reached due to the low number of drivers.

Driving simulators are increasingly important in vehicle development, benefiting from hardware and motion cueing algorithm (MCA) advancements. However, current state-of-the-art MCAs are optimising with regards to a vehicle fixed vestibular system, ignoring active and passive head movements during manoeuvres. Research shows that drivers actively move their heads to focus their gaze and passively stabilising it during involuntary trunk movements, resulting in significant differences between vehicle and head yaw angles. Humans isolate trunk and head movement in the range of 0.1–1.0 Hz, suggesting neck-driven gaze stabilisation. This behaviour is not accounted for in current MCAs, warranting an investigation. This study develops a driver gaze model to enhance motion cueing strategies and compares it to existing methods. Findings indicate significant discrepancies between vehicle and estimated head yaw rate in winter testing with high slip angles, and that omitting vestibular models and separating the slip angle in the yaw feedback improves the motion cueing with regards to induced head movements. Further, the results shows that there is a clear relationship between motion cueing, visual feedback, and induced driver head movement. In conclusion, driver gaze models can improve motion cueing strategies in driving simulators, and thus the study highlights the need for considering driver gaze behaviour and provides insights for tuning and optimising MCAs.

Subjective vehicle stability evaluation is generally conducted during closed-loop driving in which the driver controls the vehicle through transient manoeuvres and evaluate how the vehicle responds to steering inputs, especially how the slip angle builds up. To conduct such evaluation in moving base, driving simulators require that the fed vehicle response to steering inputs is representative. Limited simulator workspace often requires motion scaling, introducing errors in planar dynamics This work, therefore, investigates how the scaling of the planar motion and the slip filtering should be performed in a driving simulator, including the relative relationships between lateral acceleration, yaw rate and slip rate. Two strategies were developed based on the scaling of planar circular motion: one retaining radius information, and the other retaining velocity information. Both strategies avoided filtering the slip rate, as simulations show that the slip rate should be separated from the high-pass filtering process in cueing algorithms and that the scaling should be equal to that of the yaw rate to avoid false cues. A subjective assessment was conducted, and the results indicate advantages for retaining radius information and conclusively the advantage of unfiltered slip information in motion cueing for stability evaluation.

This paper presents a novel approach to motion cueing in moving base driving simulators (MBDS) for early-phase vehicle development. Traditional motion cueing algorithms (MCAs) rely on high-pass filtering of global vehicle states, which can reduce realism and limit the effectiveness of simulation-based subjective evaluations. The proposed method partitions vehicle motion into two components: terrain-induced (global) and vehicle-induced (local) motion. This allows for selective filtering, preserving low-amplitude high frequency vehicle responses while attenuating low frequency high amplitude terrain effects which does not fit in simulator workspaces. Two partitioning strategies are explored: a velocity-based method for pitch motion and a more general ground-plane estimation using tyre contact points. Simulation results demonstrate improved correspondence between actual and simulated pitch and roll motion, particularly in vehicle-induced motion such as braking or turning. The approach shows potential to enhance motion feedback without exceeding platform limits, enabling more effective virtual prototyping and subjective driver feedback early in the design process. Future work includes implementation in real simulators. The method supports sustainable vehicle development by reducing reliance on physical prototypes and accelerating iterative design.

Driving simulators depend on motion-cueing algorithms (MCAs) to recreate the motion cues a driver would experience in a real vehicle. Many modern MCAs, especially MPC-based approaches, incorporate simplified vestibular models and assume the driver’s head is rigidly fixed to the simulator cabin. This assumption contradicts known head-neck dynamics: the vestibulo-collic reflex (VCR) and vestibulo-ocular reflex (VOR) stabilise the head and gaze relative to the external world across a broad frequency range, meaning the actual vestibular input during driving can differ significantly from that predicted by a fixed-head model.This work investigates how accounting for basic VCR-like behaviour changes the pre-dicted semicircular canal (SCC) response, and how different cueing strategies influence gaze behaviour and retinal slip. Three strategies are compared: a classical washout filter, a model predictive control (MPC) approach using a fixed vestibular model, and a state-partitioned method that isolates global terrain-induced motion from vehicle-body-related pitch. A full vehicle model, a simple head-neck model for the VCR, and a standard SCC model are used to simulate a crest-and-brake manoeuvre representative of typical pitch dynamics. The results show that while MPC performs well for low-frequency terrain tran-sitions, it introduces a notable phase delay during braking due to filtering and the structure of the vestibular reference. This delay produces higher retinal slip during pitch onset. The classical washout algorithm provides consistently poor correlation with the vehicle motion. The state-partitioned approach tracks the vehicle most accurately, preserves the intended relationship between platform pitch and gaze, and maintains the lowest retinal slip across the manoeuvre. The comparison between fixed and dynamic SCC predictions further high-lights that fixed-head vestibular assumptions underestimate true vestibular input when lon-gitudinal acceleration induces head pitch. The findings indicate that improving simulator-vehicle motion correlation is more beneficial than relying on fixed vestibular models within the MCA.

This paper investigates if the vehicle performance envelope could be used to reduce false cues for motion cueing where the future trajectory of the vehicle is unknown. The use of driving simulators for performance driving and vehicle chassis testing is increasing due to improvement in simulator hardware and motion cueing, e.g., MPC based motion cueing, optimising the simulator motion over a future horizon. For vehicle chassis testing purposes, where the future trajectory is not necessarily well known, the conventional method is to move the platform back to neutral position, causing false cues and less efficient use of the workspace. The proposed method relies on tracking a reference position based on current vehicle acceleration and reachable accelerations. Results show better reference tracking, more efficient use of the workspace and reduced false cues.

The significance of driving simulators in vehicle dynamics development has increased during recent years in order to reduce for example development time and expensive real-life tests. A key component in driver-in-the-loop simulators is the motion cueing which plays an important role in the perception of vehicle characteristics. Motion cueing, traditionally limited by platform constraints and bandwidth, has seen advancements through algorithms like the classical washout algorithm and linear quadratic optimal control regulators (LQR), which are hand-tuned to provide effective feedback within simulator constraints. However, these methods do not inherently address workspace limitations. Further, they are commonly combined with washout filters to centre the simulator platform after an onset cue, increasing false cues. Model predictive control (MPC) has lately been used to over-come these problems, however, MPC requires significantly higher computational power and is challenging to tune and run with nonconvex workspaces. This work therefore presents a novel method for objective motion cueing tuning by considering vehicle capacity and workspace constraints together with the motion cueing. The method utilises the performance envelope (acceleration envelope) of the vehicle as constraints for the input into the motion cueing algorithm which enables forward reachability analysis. The reachable set is then adapted offline to the workspace. The results show that the algo-rithm can effectively adapt the motion cueing to the available workspace. This allows for optimising the gain and workspace offsets for filter based cueing strategies, avoiding washout to neutral and computational heavy cueing such as MPC, while considering workspace constraints and enabling both velocity and manoeuvre dependent positioning. The results show that the method improves the workspace utilisation compared to the naive approach to simply tuning the simulator gains, and by making the reachable set velocity dependent it is possible to further maximise the tuning gains.