Numerical Analysis of Tractor Accidents using Driving Simulator for Autonomous Driving Tractor

<abstract> <b>Abstract.</b> It is estimated that over 100 farmers die every year in South Korea because of accidents related to agricultural machinery. The National Institute of Agricultural Science (NAS) has conducted periodic surveys to provide basic data for the establishment of prevention strategy. Based on the analysis of the surveyed accident cases, a risk assessment of hazards causing accidents, a new approach, was conducted. A risk assessment is generally done through 4 major processes; hazards identification, frequency estimation, estimation of consequence severity, and risk evaluation. In this paper, the results of formal 2 processes for tractor accidents were introduced. Hazards were identified by reviewing the accident case records and then confirmed by the results of Delphi survey, of which participants were 27 experts on agricultural machinery accidents. The type of each accident cases were also classified during the case reviews because the estimations of frequency and consequence severity were carried out separately by types. Frequencies of hazards were estimated by following process: (1) estimation of probability of each hazards in each type, (2) calculation of frequencies of each type, and (3) frequency estimations of each hazards. Among 588 accident cases collected by NAS, 215 cases were farm-work accidents and the other 373 were traffic accidents. By the kind of a machine, the number of two-wheel tractor accidents was 453 cases, of which 178 was farm-work accident and 275 was traffic accident while that of four-wheel tractor was 135, of which 37 was farm-work accident and 98 was traffic accident. Results of accident records review and Delphi survey identified 56 hazards of tractor accidents and the accident cases were classified into 19 types, 10 for farm-work accidents and 9 for traffic accidents. The most frequent type of farm-work accidents was rollover in road and that of traffic accidents was rear collision with no fault of a tractor operator while travelling straight. The results of frequency estimations of all 56 hazards showed that carelessness was the most frequently contributing hazard followed by negligence of front watching in both two-wheel and four-wheel tractors.

Annually, tractor accidents are estimated to account for more than 100 deaths in South Korea. Periodic accident surveys have served as an essential means for the National Institute of Agricultural Sciences (NAS) to develop strategies to prevent tractor accidents. In this study, hazards leading to accidents were identified, and their risks were assessed based on survey results to establish a more effective accident prevention strategy. Risk assessment for hazards proceeded as follows: hazard identification, frequency estimation, number of equivalent fatalities (NEF) estimation, and finally risk evaluation. Hazards were identified by analyzing 588 accident cases from NAS surveys and performing an expert review of the analysis results by implementing a Delphi survey. The frequency and NEF of each hazard were estimated by multiplying its probabilities and the statistical results of the NAS surveys. Each hazard was plotted in a frequency-NEF (FN) diagram and evaluated according to its position. Fifty-four hazards were identified, and their frequencies and NEF values were estimated. The risk evaluation results, based on the FN diagram, revealed that no hazard was located in the "unacceptable" area, and two hazards (carelessness and not looking ahead carefully) were in the "as low as reasonably practicable" area. Thus, it is critical to mitigate the effects of these two hazards. With the risk assessment method used in this study, personnel who are engaged in the prevention of tractor accidents, such as policymakers, extension specialists, and researchers, can quantitatively predict how many cases or fatalities can be reduced by eliminating a certain hazard.

In recent decades, many attempts have been made to establish the cause-effect relationship model of accidents, while little work has been carried out to comprehensively consider the interdependence between the causal factors and their complex interactions with the accident outcomes. In this study, a novel accident analysis approach based on Bayesian networks (BNs) was proposed to achieve quantitative accident analysis and dynamic risk prediction of accident types and consequences. To develop the BN-based accident analysis model, a total of 1144 accident cases occurred in tank farm of China from 1960 to 2018 were collected. The BN model that can comprehensively characterize the dependencies among accident elements was established through structural learning based on accident case analysis and parameter learning based on EM algorithm. The reliability and validity of the BN model were verified by k-fold cross-validation method and comparison of predicted data with real data, and the results showed that the BN model had good classification and prediction performance. Furthermore, the established BN model was applied to the accident occurred in Huangdao, China. The analysis results show that not only the accident outcome can be accurately predicted, but also the hidden correlation can be deeply explored through the established BN model. The proposed method and findings can provide technical reference for accident investigation and analysis, and provide decision support for accident prevention and risk management.

Analysis of fatal accidents with tractors in the Centre of Portugal: Ten years analysis

The live-line works are very dangerous because of direct contacts with the distribution line or neighboring contacts. So the purpose of this study is to identify the risk factor by accident occurrence form and accident case analysis, and to suggest the quantified risk index by risk occurrence frequency and risk strength analysis. And the risk index assessment is researched by accident cases analysis on work type. Accident cases of transmission distribution line are researched based on data of the Ministry of Employment and Labor in the last ten-year period (2000~2009). In results of this paper, high risk isn't always a priority of safety measures. Risk occurrence frequency and risk strength have to be considered according to detail work types, work methods and conditions of field work. And safety management measures must be planned according to risk occurrence frequency and risk strength.

Autonomous driving (AV) technology has elicited discussion on social dilemmas where trade-offs between individual preferences, social norms, and collective interests may impact road safety and efficiency. In this study, we aim to identify whether social dilemmas exist in AVs’ sequential decision making, which we call “sequential driving dilemmas” (SDDs). Identifying SDDs in traffic scenarios can help policymakers and AV manufacturers better understand under what circumstances SDDs arise and how to design rewards that incentivize AVs to avoid SDDs, ultimately benefiting society as a whole. To achieve this, we leverage a social learning framework, where AVs learn through interactions with random opponents, to analyze their policy learning when facing SDDs. We conduct numerical experiments on two fundamental traffic scenarios: an unsignalized intersection and a highway. We find that SDDs exist for AVs at intersections, but not on highways.

Geographical information systems aided traffic accident analysis system case study: city of Afyonkarahisar

All of farm fatal-injured accidents in 4 years during 1983 to 1990 occurred in Hokkaido were surveyed by interview. The results of survey is divided two report, named part 2 and 3. The details of tractor over turn accidents on the field, catch-in to machine, run over by vehicle, pinch in to machine, fall down from high position and traffic troubles were described in this report.1) The fatal-injured accidents with tractor in farm field were classified under the tumble down from field, over turn in the field and slip down at the field entrance slop. And the main reasons of accidents were described.2) The correlational chart of reasons about tractor accidents in the field showed that the most of critical condition of farm area cannot improve, and if the person who has weak point to operate farm machines alive, no man can change to him. So the operator must control the machine with fully cautions to the critical condition every time.3) The main reasons of catch in accidents to machine, run over accidents by vehicle and pinch in accidents to machine were clarified.4) The real situations that the fatal-injured accidents brings critical conditions to his family and his farm were described.

The uncertainty and complexity of autonomous driving make Deep Reinforcement Learning (DRL) appealing. DRL can optimize the expected reward by interacting with environments. However, DRL has no guarantee of safety, which is essential for autonomous driving. Besides, the reward shaping for DRL in autonomous driving is also challenging. In this work, we introduced a policy-guided trajectory planner and proposed a hierarchical structure to try to solve these problems. The high-level DRL agent’s output is policies that are considered as suggestions for the low-level policy-guided trajectory planner. To achieve the guarantee of safety, we first translated the traffic rules and policies into formal specifications. Notice that, in many cases, the policy may not be possible to be fully applied. Given certain formal specifications, we constructed a minimum-violation motion planning problem for the low-level policy-guided trajectory planner. Through this hierarchical structure, the long-term uncertainty is handled by the high-level DRL agent, and the safety is guaranteed by the low-level planner. Furthermore, the reward includes not only the violation of traffic rules but also the violation of policies. By adding the violation of policies returned by the low-level planner to the reward, the agent could explicitly learn if a policy is supported by environments and vehicle dynamics. Our method was proven to be effective through numerical experiments.

Autonomous driving has received enormous attention from the academic and industrial communities. However, achieving full driving autonomy is not a trivial task, because of the complex and dynamic driving environment. Perception ability is a tough challenge for autonomous driving, while 3D object detection serves as a breakthrough for providing precise and dependable 3D geometric information. Inspired by practical driving experiences of human experts, a pure visual scheme takes sufficient responsibility for safe and stable autonomous driving. In this paper, we proposed an anchor-free and keypoint-based 3D object detector with monocular vision, named Keypoint3D. We creatively leveraged 2D projected points from 3D objects’ geometric centers as keypoints for object modeling. Additionally, for precise keypoints positioning, we utilized a novel self-adapting ellipse Gaussian filter (saEGF) on heatmaps, considering different objects’ shapes. We tried different variations of DLA-34 backbone and proposed a semi-aggregation DLA-34 (SADLA-34) network, which pruned the redundant aggregation branch but achieved better performance. Keypoint3D regressed the yaw angle in a Euclidean space, which resulted in a closed mathematical space avoiding singularities. Numerous experiments on the KITTI dataset for a moderate level have proven that Keypoint3D achieved the best speed-accuracy trade-off with an average precision of 39.1% at 18.9 FPS on 3D cars detection.

In the field of agriculture, satellite imagery has revolutionized the development of remote sensing in food production, food security and supply chain management. However, existing solutions are too limited to solve all the problems of remote sensing big AgriData completely mainly because they are not designed to permit collaboration among stakeholders to support sharing of data and farming operations in general and create useful knowledge bases. Access to real-time AgriData, real-time forecasting and tracking of physical items will significantly change farm management and operations and in combination with IoT development will lead in the autonomous operation of farm. Our focus is to identify, explore and exploit the added value the big remote sensing AgriData provide in food security context. In this sense the main application related to cropland mapping context are also reviewed and discussed concentrating on their suitability in mapping crop types at small-scale farms.

Drinking locations of drink-drivers: A comparative analysis of accident and nonaccident cases

According to the statistics, occupational fatal injuries by m obile cranes were about 12 per year in whole industrial. Mobile cranes are widely used in various part s of industries to improve the efficiency of the work. However considerable number of fatal injuries happ en each year during the operation of the machines. In this study, the current regulations to be adeq uate in industrial site have to be renew in order to prevent the fatal injuries by mobile cranes. Fatal injury analyses were conducted with several accident ca ses by the mobile cranes. For each accident, the causes of the injuries were examined and proper s afety measures were proposed. In this study, the mobile crane showed a high fatality rate in industrial accidents and no detailed cause analysis of fatal accidents was conducted in terms of unsafe ac ts or conditions. This study proposed a revision of the standard guideline as an accident prevention me asures through in-depth analysis of fatal accidents. First, among the mainly five machines caused the accidents, m obile crane was higher for the second showed 0.6% for number of fatalities compared to number of mobi l cranes and for the third showed 11% for number of fatalities compared to number of injuries. Se cond, main cause of cognitive engineering agenda was visibility, responsibility, affordance. As the measures to prevent accidents before starting operation, alternative revision for the fool pr oof including visibility, responsibility, affordance etc. for the fool proof measures was proposed. Third , alternative revision as cognitive accident prevention for the fail safe measures was proposed.Key words : mobile crane, human error, visibility, compatibilit y, fool proof, fail safe.

This paper describes a novel manoeuvre planning method to attenuate disturbances acting on occupants of autonomous cars as a result of driving behaviour. New research findings suggested that the passengers in autonomous cars might be more prone to motion sickness and thus overall discomfort. The proposed approach is based on a recently developed novel continuous B-spline path smoothing algorithm for car-like steered robots. Two algorithms are designed for urban driving scenarios and steering between two predefined poses. The resulting paths avoid abrupt changes in steering and longitudinal velocity, by maintaining curvature and its high order continuity. We show that this lead to reduced high frequency disturbances in steering and resulting load disturbances on passengers. The presented novel B-spline manoeuvres outperform other planning methods by reducing lateral acceleration and yaw disturbances. New approach was verified by rigorous simulations, numerical and field experimentation. Tests were repeated for a number of different paths and velocities. The reported results are the first spline based parameterisation methods practically applied for autonomous cars planning and re-planning, then validated using both noisy actuation simulations and field experiments.

An Accident Causation Analysis and Taxonomy (ACAT) model of complex industrial system from both system safety and control theory perspectives