Loading
Life sciences · Preprint
arXiv · August 18, 2026
Early or partial results. Treat as a signal, not a conclusion.
This is a proof-of-concept study of a machine-learning model trained on simulator data to predict collision risk in powered two-wheeler riders under time pressure. The model achieves high accuracy metrics in the training domain (94.97% accuracy, 99.33% ROC AUC) but has not been validated in real-world settings, clinically tested, or peer reviewed. The work is exploratory and does not yet support clinical or operational deployment.
Simulator-based observational cohort study with machine-learning model development. 51 participants performing powered two-wheeler simulator rides; specific inclusion/exclusion criteria, age, experience level, and demographics not reported.. Intervention: Simulator-based powered two-wheeler rides under three time-pressure conditions (No, Low, High TP) with capture of 64 multivariate features per sequence.. Compared with: Ten baseline machine-learning models (TimesNet, LLM4TS, Time-LLM, iTransformer, and others not named) evaluated on the same tasks.. n = 51. Not reported..
MotoSafety achieved 94.97% accuracy and 99.33% ROC AUC on collision risk classification task Forecasting error 0.039 MSE and 0.094 MAE, reported as 4.4× lower than Time-LLM and iTransformer Model deployed with 1.15M parameters and 0.135 ms latency on CPU hardware
No comparison to established two-wheeler safety risk scores or clinical decision tools. MotoSafety achieved 94.97% accuracy and 99.33% ROC AUC on collision risk classification task
This model is not yet suitable for clinical or real-world safety deployment. Validation in actual riding environments, comparison to human performance, and peer review are essential before any operational use.
A simulator-based feasibility study introducing a novel machine-learning architecture for collision risk prediction in a safety domain, with no clinical or real-world validation, no peer review, and no comparison to established clinical risk stratification.
As stated by the source record.
Quoted from the source exactly as published.
This model is not yet suitable for clinical or real-world safety deployment. Validation in actual riding environments, comparison to human performance, and peer review are essential before any operational use.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Powered two-wheeler riders face critical safety challenges in low- and middle-income countries, yet limited studies exist on how cognitive stressors such as Time Pressure influence collision risk. To address this gap, we introduce a large-scale dataset of over 129,000 labeled multivariate time-series sequences from 153 simulator rides by 51 participants under No, Low, and High TP, capturing 64 features across vehicle dynamics, control inputs, proximity, and behavioral violations. Building on this dataset, we propose MotoSafety, a novel edge-AI architecture grounded in the Learned Temporal Importance principle. MotoSafety achieves 94.97% accuracy and 99.33% ROC AUC, outperforming ten baselines, including TimesNet and LLM4TS, and achieves 0.039 MSE and 0.094 MAE for forecasting (4.4x lower error than Time-LLM and iTransformer). With only 1.15M parameters and 0.135 ms latency, it is suitable for edge deployment on low-cost CPU hardware. Using ground truth TP as an inductive bias improves accuracy from 94.09% to 94.97%, while predicted TP achieves 94.82%. Using only 21 IMU+GPS features, it achieves 93.91% accuracy, indicating practical deployment. Beyond PTW safety, the architecture shows better transferability to human activity (97.66%) and clinical (99.65%) domains. This lightweight framework advances PTW collision risk assessment, supporting the Safe System Approach for Intelligent Transportation Systems.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.