Dataset of stochastic human body model simulations in frontal collisions

General remarks

These are the simulation results supplementing the PhD thesis of Felix Ressi (DOI: 10.3217/7bge8-ffb75). A detailed description of the simulations and subsequent injury analysis can be found there. The simulations were performed using a modified version of the generic vehicle interior developed by Johan Iraeus. The original model can be found here at openvt.eu. In addition, four detailed human body models were used:

• THUMS v4.1 5th percentile female
• THUMS v4.1 50th percentile male
• VIVA+ 50th percentile female
• VIVA+ 50th percentile male

The THUMS models are available free of charge from Toyota and the open source VIVA+ models are available at openvt.eu. The specific VIVA+ version used for these simulations can be found in this branch.

The input for the simulations were identical for each HBM, apart from the seat position. In the conventional driving (dynamic driving task - DDT) position, each model was positioned based on a regression model. To add some scatter to the resulting seat position, it was varied between 0 mm and 25 mm behind the predicted position. For the autonomous driving (AD) position, the predicted longitudinal seat position of the VIVA+ 50th percentile male was used as a baseline for all HBMs, from which the seat was moved rearwards between 150 mm and 250 mm. Hence, in the AD simulations, all HBMs were in identical seat positions longitudinally. All input parameters for the 200 simulations, which were performed with each HBM in both seat configurations (resulting in 1600 simulations overall), are provided in the simulation_matrix.csv. Based on the value in "Seat position factor" [0, 1], the seat position for the DDT posture [0-25 mm] (relative to each HBMs driving posture) or AD posture [150-250 mm] (realtive to the DDT seat position for the VIVA+ 50M) was calculated.

Criteria results

• DDT position - dynamic driving task (i.e. conventional driving) position
• AD - autonomous driving position (conventional seat back angle, but seat moved rearwards between 150 mm and 250 mm)

The "results" dataframes consist of 200 rows (one for each simulation variant) and 11177 columns, with the DDT data using IDs 1-200 and the AD data IDs 201-400 (facilitating potential merging of the dataframes).

The data can be read into a `pandas` dataframe by using the following line:

df = pd.read_csv("results_DDT_position.csv", header=[0,1], index_col=0)

This creates a MultiIndex column dataframe, which holds the data for all four HBMs used in the simulations. They are abbreviated in the following way:

• T05F: THUMS v4.1 5th percentile female
• T50M: THUMS v4.1 50th percentile male
• V50F: VIVA+ 50th percentile female
• V50M: VIVA+ 50th percentile male

In addition, the columns with the header "prob_of_occ" (i.e. probability of occurrence) provide information on the relative probability of occurrence of each variant based on the crash database analysis. Aside from the overall relative probability of occurrence for females and males (p_f and p_m respectively), the relative probability of occurrence  for females and males are also provided for the vehicle mass, delta-v, PDOF, and accident type individually (p_mass, p_dV, p_PDOF and p_F2x respectively).

In order to access only data of one HBM, the following line can be used (example using VIVA+ 50F data):

df_V50F = df['V50F']
display(df_V50F)

The following table lists examples of the 4339 unique criteria in the dataframes. However, as most are self-explanatory, only potentially ambiguous ones are listed.

[Table coming soon...]

Kinematics results

The eight files contain the kinematics data for each HBM in each of the two positions separately. The dataframes are organized by columns, where each columns represents a simulation variant, which in turn are divided into "VEHICLE", "HEAD", and "RELATIVE". The rows represent time steps, which means that the 1500 rows represent the 150 ms simulation time in a 0.1 ms interval. 

To read these files into a `pandas` dataframe, use the following code (example using VIVA+ 50F data):

df_kin_V50F = pd.read_csv("kinematics_V50F_DDT_position.csv", header=[0,1,2,3,4], index_col=0)

Simulation animations

Corresponding animations for each of the 200 simulations in side (left and right), top and rear view are provided under separate DOIs for each HBM (due to the upload size limit).

• DDT position with THUMS v4.1 (5th percentile female and 50th percentile male): 
• AD position with THUMS v4.1 (5th percentile female and 50th percentile male): 
• DDT position with VIVA+ 1.0.3-alpha (50th percentile female and 50th percentile male): 10.3217/5spba-g0t54
• AD position with VIVA+ v1.0.3-alpha (50th percentile female and 50th percentile male): 10.3217/b9txy-k7663

Important note for statistical analyses

Please note that due to an oversight in mapping the Latin hypercube design (where all variables use values between zero and one) to the simulation input parameters (where the variables use a specified parametric distribution and range) the variables "PDOF" and the seatbelt load limiter level are perfectly correlated. However, the actual values are fine, this issue only affects statistical analyses, where this correlation can lead to errors. The parameter correlation is illustrated by the simulation matrix provided in the simulation_matrix.png.