Jared Gragg

Hybrid method for driver accommodation using optimization-based digital human models

Many applications exist where humans are required to perform a task in a seated position, such as operating a vehicle. Seated posture inside a vehicle influences driver performance and control of the vehicle. For people of extreme stature, tall or short, and for people of extreme width, obese or pregnant populations, it can be difficult to safely operate a vehicle if there is not enough room in the cab or if some controls cannot be reached. This study proposes a hybrid method for predicting the optimum driver seat adjustment range to accommodate diverse drivers in any vehicle based on a direct optimization-based posture prediction method. The proposed hybrid method combines three approaches: a boundary manikin approach, a population sampling approach, and a special population approach. The boundary manikin approach places two boundary manikins (5% female and 95% male) inside a virtual vehicle cab to perform tests. The population sampling approach spans a multitude of test subjects ranging in stature from 158 to 185 cm, determining the range from a plot of predicted hip point distances from the point of contact of the right heel and the floor. The special population approach studies the effect that size and shape changes, such as pregnancy, have on seated posture inside a vehicle. Also given is an indication of discomfort through the output values of the multi-objective function in the optimization formulation. A combination of the three approaches is used to determine an optimal adjustment range for the driver seat, thus allowing most people to safely operate the vehicle. Results of the simulations are validated using experimental determinations of the driver seat adjustment range from the literature. The main benefit of using this method is that the human aspect of design can be included early in the design process, thereby reducing or eliminating prototypes. Another benefit of the simulation is that it can be adapted to other seating tasks such as: occupant seat check inside a vehicle; workstation design; and issues related to other special populations such as obese individuals, dwarfs, and children.

Effect of human link length determination on posture reconstruction

Motion capture experiment results are often used as a means of validation for digital human simulations. Motion capture results are marker positions and joint centers in Cartesian space. However, joint angles are more intuitive and easy to understand compared to marker or joint center positions. Posture reconstruction algorithms are used to map Cartesian space to joint space by re-creating experimental postures with simulation models. This allows for direct comparison between the experimental results and digital human simulations. Besides the inherent experimental errors from motion capture system, one source of simulation error is the determination of the link lengths to be used in the simulation model. The link length errors can propagate through all simulation results. Therefore, it is critical to eliminate the link length errors. The objective of this study is to determine the best method of determining link lengths for the simulation model to best match the model to the experiment results containing errors. Specifically, the way that the link lengths are calculated in the posture reconstruction process from motion capture data has a significant effect on the recreated posture for the simulation model. Three link length calculation methods (experimental-average method, trial-specific method, and T-pose method) are developed and compared to a benchmark method (frame-specific method) for calculating link lengths. The results indicate that using the trial-specific method is the most accurate method when referring to calculating frame-specific link lengths.

An inverse optimization approach for determining weights of joint displacement objective function for upper body kinematic posture prediction

Human posture prediction can often be formulated as a nonlinear multiobjective optimization (MOO) problem. The joint displacement function is considered as a benchmark of human performance measures. When the joint displacement function is used as the objective function, posture prediction is a MOO problem. The weighted-sum method is commonly used to find a Pareto solution of this MOO problem. Within the joint displacement function, the relative value of the weights associated with each joint represents the relative importance of that joint. Usually, weights are determined by trial and error approaches. This paper presents a systematic approach via an inverse optimization approach to determine the weights for the joint displacement function in posture prediction. This inverse optimization problem can be formulated as a bi-level optimization problem. The design variables are joint angles and weights. The cost function is the summation of the differences between two set of joint angles (the design variables and the realistic posture). Constraints include (1) normalized weights within limits and (2) an inner optimization problem to solve for joint angles (predicted posture). Additional constraints such as weight limits and weight linear equality constraints, obtained through observations, are also implemented in the formulation to test the method. A 24 degree of freedom human upper body model is used to study the formulation and visualize the prediction. An in-house motion capture system is used to obtain the realistic posture. Four different percentiles of subjects are selected to run the experiment. The set of weights for the general seated posture prediction is obtained by averaging all weights for all subjects and all tasks. On the basis of obtained set of weights, the predicted postures match the experimental results well.

An alternative formulation for determining weights of joint displacement objective function in seated posture prediction

The human posture prediction model is one of the most important and fundamental components in digital human models. The direct optimization-based method has recently gained more attention due to its ability to give greater insights, compared to other approaches, as how and why humans assume a certain pose. However, one longstanding problem of this method is how to determine the cost function weights in the optimization formulation. This paper presents an alternative formulation based on our previous inverse optimization approach. The cost function contains two components. The first is the weighted summation of the difference between experimental joint angles and neutral posture, and the second is the weighted summation of the difference between predicted joint angles and the neutral posture. The final objective function is then the difference of these two components. Constraints include (1) normalized weights within limits; (2) an inner optimization problem to solve for the joint angles, where joint displacement is the objective function; (3) the end-effector reaches the target point; and (4) the joint angles are within their limits. Furthermore, weight limits and linear weight constraints determined through observation are implemented. A 24 degree of freedom (DOF) human upper body model is used to study the formulation. An in-house motion capture system is used to obtain the realistic posture. Four different percentiles of subjects are selected and a total of 18 target points are designed for this experiment. The results show that using the new objective function in this alternative formulation can greatly improve the accuracy of the predicted posture.

Posture reconstruction method for mapping joint angles of motion capture experiments to simulation models

Motion capture experiments are often used in coordination with digital human modeling to offer insight into the simulation of real-world tasks or as a means of validating existing simulations. However, there is a gap between the motion capture experiments and the simulation models, because the motion capture system is based on Cartesian space while the simulation models are based on joint space. This paper bridges the gap and presents a methodology that enables one to map joint angles of motion capture experiments to simulation models in order to obtain the same posture. The posture reconstruction method is an optimization-based approach where the cost function is a constant and constraints include (1) the distances between simulation model joint centers and the corresponding experimental subject joint centers are equal to zeros; (2) all joint angles are within joint limits. Examples are used to demonstrate the effectiveness of the proposed method.

Optimisation-based approach for determining driver seat adjustment range for vehicles

When designing for human variability in the interior cab design for a vehicle, it becomes difficult to predict an optimum driver seat adjustment range. This paper proposes an optimisation–based approach to determine the seat adjustment range without the need for population sampling and stochastic posture prediction. This paper uses boundary anthropometric digital human models, a 95% male and a 5% female, to establish the driver seat adjustment range for vehicles. The simulation predicts the optimum posture of the seated driver inside the vehicle, and also gives an indication of how comfortable the driver is while seated in the predicted posture.

Optimization-based posture reconstruction for digital human models

Digital human modeling provides a valuable tool for designers when implemented early in the design process. Motion capture experiments offer a means of validation of the digital human simulation models. However, there is a gap between the motion capture experiments and the simulation models, as the motion capture results are marker positions in Cartesian space and the simulation model is based on joint space. Therefore, it is necessary to map the motion capture data to simulation models by employing a posture reconstruction algorithm. Posture reconstruction is an inherently redundant problem where the collective distance error between experimental joint centers and simulation joint centers is minimized. This paper presents an optimization-based method for determining an accurate and efficient solution to the posture reconstruction problem. The procedure is used to recreate 120 experimental postures. For each posture, the algorithm minimizes the distance between the simulation model joint centers and the corresponding experimental subject joint centers which is called the mean measurement error.

Predicting the probability of slip in gait: methodology and distribution study

The likelihood of a slip is related to the available and required friction for a certain activity, here gait. Classical slip and fall analysis presumed that a walking surface was safe if the difference between the mean available and required friction coefficients exceeded a certain threshold. Previous research was dedicated to reformulating the classical slip and fall theory to include the stochastic variation of the available and required friction when predicting the probability of slip in gait. However, when predicting the probability of a slip, previous researchers have either ignored the variation in the required friction or assumed the available and required friction to be normally distributed. Also, there are no published results that actually give the probability of slip for various combinations of required and available frictions. This study proposes a modification to the equation for predicting the probability of slip, reducing the previous equation from a double-integral to a more convenient single-integral form. Also, a simple numerical integration technique is provided to predict the probability of slip in gait: the trapezoidal method. The effect of the random variable distributions on the probability of slip is also studied. It is shown that both the required and available friction distributions cannot automatically be assumed as being normally distributed. The proposed methods allow for any combination of distributions for the available and required friction, and numerical results are compared to analytical solutions for an error analysis. The trapezoidal method is shown to be highly accurate and efficient. The probability of slip is also shown to be sensitive to the input distributions of the required and available friction. Lastly, a critical value for the probability of slip is proposed based on the number of steps taken by an average person in a single day.

Joint discomfort human performance measure for driver posture prediction: some insights

Digital human models have proven to be valuable tools for understanding human reach envelope inside a vehicle. Typical digital human posture prediction simulations employ optimisation techniques that find the most likely posture that a human would realise to achieve a given task. Human performance measures are included as an objective function in the optimisation formulation. A previous study (Yang et al., 2004) defined a joint discomfort human performance measure based on joint angles that incorporates three distinct aspects of joint discomfort. However, the previously defined joint discomfort function is poorly understood. This paper investigates the properties of the joint discomfort function and how each parameter in the function affects the predicted posture. An alternate formulation of the joint discomfort human performance measure is proposed. The parameters of the new joint discomfort function are investigated through graphical analysis, ANOVA analysis, and sensitivity analysis. The joint discomfort function is then employed in several posture prediction simulations that pertain to reach space inside a vehicle. The postures are given to demonstrate the effect that the parameters of joint discomfort have on predicted postures.

Probabilistic sensitivity analysis of in-vehicle reach tasks for digital human models considering anthropometric measurement uncertainty

For design using digital human models, human anthropometry data are required as input and are extracted from measurements. There is inherent error associated with these measurements which impacts the output of the simulation. Current techniques in digital human modeling applications primarily employ deterministic methods which are not well suited for handling variability in anthropometric measurement. An alternative to deterministic methods is probabilistic/sensitivity analysis. This study presents a probabilistic sensitivity approach to gain insights into how uncertainty in anthropometric measurements can affect the results of a digital human model with the specific application of vehicle-related reach tasks. Sensitivity levels are found to determine the importance of variability in each joint angle and link length to the final reach. A55-degree of freedom (DOF) digital human model is introduced to demonstrate the sensitivity approach for reach tasks. Seven right-hand reach target points and two left-hand reach target points (creating a total of 14 reach tasks) within a vehicle are used to compare the sensitivities in the joint angles and link lengths resulting from measurement uncertainty. The results show that the importance of each joint angle or link length is dependent on the characteristics of the reach task and sensitivities for joint angles, and link lengths are different for each reach task.