Responsibility analysis was conducted on almost 10,000 drivers injured over a 10-year period in Victoria, Australia as a result of a vehicular collision and in whom comprehensive toxicology testing in blood was conducted. Multivariable logistic regression models were fitted to compute unadjusted (OR) and adjusted odds ratios (aOR). In the multivariable model, adjustments were made for gender, age group, vehicle type, and whether the crash occurred in metropolitan or rural areas. The psychoactive drug-free driver was the control group. Drivers with ethanol (alcohol) present at any concentration (>0.01 % BAC) confirmed the well-known large increases in OR that occur with rising blood ethanol concentration, with an aOR of 16.4 (95 %CI 11.1–24.2). Drivers positive to methamphetamine and no other psychoactive drug also showed a large increase in the odds of culpability (aOR 17.2, 95 %CI 10.2–29.2). Drivers with Δ9-tetrahydrocannabinol (THC) in blood showed a modest increase in odds when all concentrations (> 1 ng/mL) were assessed (aOR 1.5, 95 %CI 1.1–2.0) that trended to a possible increased aOR 2.1 (95 %CI 0.9–4.9) when concentrations were 10 ng/mL or higher. Benzodiazepines in drivers also showed a significant increase in odds (aOR1.8, 95 %CI 1.2–2.8), as well as antidepressants (aOR 1.3, 95 %CI 1.0–1.6). Opioids such as codeine and oxycodone showed no measurable increase in the odds of culpability, whereas combinations of ethanol with methamphetamine or THC gave aOR of 22.6 (95 %CI 3.0–168) and 20.0 (95 %CI 7.3–54.5), respectively, as well as drivers positive to both THC and methamphetamine (aOR 14.6, 95 % CI 5.9–36.4). Drivers positive to one or more illicit drug, or a larger selection of psychoactive drugs in the absence of ethanol both gave similarly large increase in odds (aOR 4.0, 95 %CI 3.3–5.0). This study contributes to a better understanding of risks associated with driving in the presence of common psychoactive drugs.
Motorized three-wheeler rickshaws are a major urban transport mode across South Asia, yet their injury-severity patterns remain insufficiently understood. Existing rickshaw safety studies rely largely on structured crash attributes, while narrative-based approaches often treat textual information as a single undifferentiated source. As a result, the distinct roles of contextual conditions, administratively reported crash causes, and crash-event descriptions are rarely preserved within a unified and interpretable prediction framework. To address this gap, this study proposes a mechanism-oriented Distributed Crash Reasoning (DCR) framework for interpretable multiclass injury-severity prediction. DCR preserves three functionally distinct evidence streams: structured contextual conditions, administratively reported crash-cause information, and crash-event descriptions. DCR processes these evidence streams through separate reported-cause, event-interaction, and structured-context modules and integrates them through attention-based multimodal fusion. The framework was evaluated using 2,742 motorized rickshaw crash records from Rawalpindi, Pakistan, covering 2017–2019. The proposed DCR achieved the highest recall, F1-score, and PR-AUC among the evaluated models. Controlled experiments further showed that multimodal structured-text configurations generally outperformed unimodal alternatives. Evidence-stream ablation and faithfulness analyses further identified structured context and crash-event information as the principal predictive sources, while reported-cause information provided a complementary contribution, particularly for Fatal-class discrimination. The multimodal advantage also persisted when models trained on 2017–2018 records were evaluated on 2019 data, indicating later-year predictive consistency within the available dataset. The statistical analysis further identified significant temporal (i.e., weekend, peak period, and nighttime), roadway (i.e., posted speed limit), vehicle (i.e., number of vehicles involved), and human (i.e., driver age and gender) factors, with their associations differing between Severe and Fatal injury outcomes. These findings demonstrate the value of preserving the semantic roles of heterogeneous crash information for class-imbalanced injury-severity prediction while providing diagnostically traceable evidence on the relative roles of different information streams in model predictions.
Advances in automated driving technology have facilitated the commercial deployment of Level 3 systems. Complex traffic scenarios still require drivers to resume vehicle control before full automation is realized. Assessment of driver takeover performance and associated risks is essential for enhancing automated driving safety. Current literature focuses primarily on system-initiated takeovers. Research on driver-initiated takeovers remains limited. This study addressed this gap by investigating driver-initiated takeovers triggered by traffic pressure within highway tunnels. The experimental design included four non-driving-related tasks (NDRTs) and a no-task condition. Visual, physiological, and behavioral data were collected from 28 drivers across five experimental conditions using a high-fidelity driving simulator. An evaluation index system was established based on the takeover behavior from three dimensions: takeover performance, driving workload, and risk level. A quantitative grading model for driving workload level (DWL) was developed using kernel principal component analysis and fuzzy C-means clustering. Additionally, a cloud entropy-optimized extension cloud model was used to comprehensively evaluate the risk of driver-initiated takeovers. The results indicated differences in takeover performance, driving workload, and model-derived risk assessment results across the tested task conditions. Significant differences in driver-initiated takeover time were observed across task conditions, reflecting variations in the timing of driver-initiated intervention. Short-term speed compensation was generally observed after takeover and was more pronounced under the visual and multimodal task conditions. DWL values were higher under all NDRTs than under the no-task condition, with the highest value observed under the multimodal task condition. DWL was classified into three levels: low [0, 0.537), medium [0.537, 0.816), and high [0.816, 1]. Within the relative risk classification for the current sample and experimental scenario, the no-task condition was classified as low risk; the auditory, visual, and social task conditions were classified as medium risk; and the multimodal task condition was classified as high risk. The THW results provided preliminary convergent evidence for these risk classifications. These findings provide a scientific basis for takeover warnings, driver state assessment, and human–machine co-driving safety interventions in highway tunnel scenarios.
In-wheel motor driven distributed drive electric vehicles (DDEVs) extend vehicle dynamic performance boundaries with independent torque control capabilities. However, the deep control coupling between active front steering (AFS) and direct yaw control (DYC) induces functional redundancy and execution conflicts, significantly increasing instability risks, which are a critical precursor to loss-of-control accidents in driver-system shared control scenarios. To address this, this paper constructs a handling stability-oriented cooperative integrated control framework for AFS and DYC. The system parameter uncertainties are handled based on the finite-vertex polytope method, and a compact polytopic robust model is formulated through convex optimization theory. Building upon this modeling, a distributed model predictive control (DMPC)-driven multi-agent dynamic interaction architecture is designed to realize real-time state prediction and cooperative decision optimization. Furthermore, a parallel distributed control mechanism is established by incorporating cooperative game theory to coordinate dynamic weight distribution of actuators, thereby satisfying the modularity and scalability requirements of chassis integrated control. To guarantee system asymptotic stability, terminal input constraints and cost functions are incorporated into a game-theoretic formulation, which is formalized within a linear matrix inequality (LMI) solution framework. An H-infinity robust compensation mechanism is further integrated to suppress external disturbances, which plays a key role in preventing accident-prone instability under system uncertainties. Finally, validation through hardware-in-the-loop (HIL) tests and real vehicle experiments demonstrates the significant effectiveness of the proposed scheme in enhancing lateral stability, reducing accident risk, and improving active safety performance in complex driving scenarios.
Autonomous vehicles (AVs) are emerging as intelligent mechanical systems for safer transportation. However, in safety–critical scenarios, existing AVs still rely largely on rule-based, ego-centric strategies, possibly increasing collision and injury risk to other road users. Progress towards safer autonomy is constrained by the rarity of such events and the lack of explicit consideration of post-crash driver injury outcomes. We propose an integrated safety framework for learning and evaluating AV safety performance, enabling AVs to internalize predictable human driving strategies and quantify the extent of injury mitigation in unavoidable collision scenarios. By accelerating the generation of safety–critical behaviors from experienced human drivers and linking pre-crash decisions to in-crash injury through high-biofidelity collision simulation, we evaluate safety performance via human likeness, collision rate, and injury severity. Using 1,988 accelerated driver-in-the-loop collision-avoidance behaviors for model learning and injury-aware evaluation, our human-informed model emulates human-like evasive strategies and, within the experimental paradigm of this study, reduces the combined probability of AIS 3+ injury by 36.6% (95% CI −28.3% to 68.2%) and the median per-crash economic loss by approximately 23% (255,093 USD vs. 196,669 USD) in crashes that still occurred. This work demonstrates that incorporating human evasive experience into autonomous driving learning tends to reduce driver injury severity in collisions under safety–critical scenarios.
