Marinus Maria van Paassen

Active Deceleration Support in Car Following

A haptic gas pedal feedback system is developed that provides car-following information via haptic cues from the gas pedal. During normal car-following situations, the haptic feedback (HF) cues were sufficient to reduce control activity and improve car-following performance. However, in more critical following situations, drivers use the brake pedal to maintain separation with the lead vehicle. A deceleration control (DC) algorithm is designed that, in addition to the HF, provided increased deceleration upon release of the gas pedal during car-following situations that required faster deceleration than releasing the gas pedal alone would do. For the design, a driver model for car following in different situations was estimated from driving simulator data. A Monte Carlo analysis with the driver model yielded subjective decision points, where drivers released the gas pedal to start pressing the brakes. This enabled the definition of a reaction field, which determined the needed deceleration input for the DC algorithm. The tuned DC algorithm was tested in a fixed-base driving simulator experiment. It was shown that the active deceleration support improved the car-following performance while reducing the driver brake pedal input magnitude in the conditions tested.

Identification of the Feedforward Component in Manual Control With Predictable Target Signals

In the manual control of a dynamic system, the human controller (HC) often follows a visible and predictable reference path. Compared with a purely feedback control strategy, performance can be improved by making use of this knowledge of the reference. The operator could effectively introduce feedforward control in conjunction with a feedback path to compensate for errors, as hypothesized in literature. However, feedforward behavior has never been identified from experimental data, nor have the hypothesized models been validated. This paper investigates human control behavior in pursuit tracking of a predictable reference signal while being perturbed by a quasi-random multisine disturbance signal. An experiment was done in which the relative strength of the target and disturbance signals were systematically varied. The anticipated changes in control behavior were studied by means of an ARX model analysis and by fitting three parametric HC models: two different feedback models and a combined feedforward and feedback model. The ARX analysis shows that the experiment participants employed control action on both the error and the target signal. The control action on the target was similar to the inverse of the system dynamics. Model fits show that this behavior can be modeled best by the combined feedforward and feedback model.

Reducing steering wheel stiffness is beneficial in supporting evasive maneuvers

Most collision avoidance systems for highway scenarios are shifting the role of the driver from manual execution to a supervisory position. In the interface design of the proposed collision avoidance system proposed in this article, the human-machine issues complicating task automation were avoided by adopting a human-centered approach. In this paper a method to preserve drivers choice and maneuver initiation through the use of a temporarily unstable steering wheel is presented. In the particular driving condition investigated in a fixed base driving simulator, the driver has to avoid an obstacle following one of two equally safe escape paths (left or right evasive maneuver). The collision avoidance system aims at supporting both available solutions by means of a haptic steering wheel interface. Two feedback modalities are used, namely torque feedback and stiffness feedback. The results of the experiment show that the haptic interface effectively reduced the number of crashes, decreased response time with at least 100 ms while reducing the control effort and activity in the most critical situations.

Automatically Generating Specification Properties From Task Models for the Formal Verification of Human-Automation Interaction

Human-automation interaction (HAI) is often a contributor to failures in complex systems. This is frequently due to system interactions that were not anticipated by designers and analysts. Model checking is a method of formal verification analysis that automatically proves whether or not a formal system model adheres to desirable specification properties. Task analytic models can be included in formal system models to allow HAI to be evaluated with model checking. However, previous work in this area has required analysts to manually formulate the properties to check. Such a practice can be prone to analyst error and oversight which can result in unexpected dangerous HAI conditions not being discovered. To address this, this paper presents a method for automatically generating specification properties from task models that enables analysts to use formal verification to check for system HAI problems they may not have anticipated. This paper describes the design and implementation of the method. An example (a pilot performing a before landing checklist) is presented to illustrate its utility. Limitations of this approach and future research directions are discussed.

The importance of including knowledge of neuromuscular behaviour in haptic shared control

Haptic shared control is a powerful way of combining the best of humans and intelligent vehicles, keeping humans in the loop while avoiding many automation issues. Literature has shown that haptic shared control can support drivers to increase performance at reduced control effort, but also points out that even then, subtle conflicts occur between driver and shared controller. This paper hypothesizes that at least part of that disagreement lies at the neuromuscular level, and that mismatches in expected torques will result in decreased performance and increased effort. The goal of this paper is to provide experimental evidence that shows the importance of tuning guidance torques with the correct expectations about neuromuscular response. An abstract steering experiment was performed without visual cues, where drivers were guided by haptic shared control torques to perform a lane-change maneuver. The torques were tuned with three different expectations about drivers neuromuscular behavior, and drivers were also instructed to perform three different neuromuscular tasks. The results show that when the tuning of the torques did not match the real neuromuscular behavior, guidance torques were either too high or too low, and performance was reduced. It is concluded that a good understanding of neuromuscular response of drivers is essential to avoid subtle conflicts between driver and shared controller.

Biodynamic feedthrough is task dependent

Vehicle accelerations may lead to involuntary limb motions. These motions can result into involuntary control inputs when performing a manual control task. This phenomenon is called biodynamic feedthrough (BDFT). This paper aims to show that task interpretation plays an important role in the occurrence of BDFT. Results of an experiment are presented, in which biodynamic feedthrough was measured during three different control tasks. Each control task required the human operator to adapt his/her neuromuscular settings. The results show that the level of biodynamic feedthrough depends on the task the human operator is performing. From further analysis, it can be observed that the experiment results are in good agreement with BDFT measurements found in literature. The comparison confirms that the task interpretation plays an important role in BDFT which cannot be ignored when attempting to understand or mitigate BDFT in practical situations.

Aerodynamic Hinge Moment Coefficient Estimation Using Automatic Fly-by-Wire Control Inputs

This paper describes the modeling and parameter estimation of the aileron and elevator flight control system of TU Delft’s Cessna Citation II laboratory aircraft. The flight test data originate from maneuvers performed autonomously with a custom-designed experimental fly-by-wire system. The identification of the aerodynamic hinge moment coefficients will be of special interest, as these hinge moments greatly affect the in-flight performance of the flight control system. First, the elevator and aileron flight control system models will be presented, introducing the main parameters that need to be determined. Most of the parameters reflect the mechanical properties and can be obtained through some cleverly-designed ground tests, which are discussed next. The hinge moment coefficients can only be determined through flight tests. The paper continues with a description of the optimal input signals used to generate flight data for the parameter estimation procedure. The flight test setup will be introduced briefly, after which the results of the hinge moment coefficient parameter estimation are summarized. Finally, the validity of the resulting flight control system models for elevator and aileron are shown.

A Machine Learning Approach to Trajectory Prediction

A machine learning approach to trajectory prediction for sequencing and merging of traffic following fixed arrival routes is described and evaluated using actual aircraft trajectory and meteorological data. In the approach a model is trained using historic data to make arrival time predictions. Model inputs are the aircraft type, aircraft ground speed and altitude at the start of the arrival route, surface wind, and altitude winds. A stepwise regression method is used to systematically determine the inputs and functions of inputs that are included in the prediction model based on their explanatory power. For the evaluation of the approach a 45 NM fixed arrival route was used that ends at the runway. Traffic performed a continuous descent operation. At a prediction horizon of 45 NM the model explained 63% of the observed variance in the arrival time. The mean absolute time error was 18 s. Finally, the prediction model was used to determine the required initial spacing interval between aircraft for continuous descent operation and examine the impact on runway throughput and conflicts. Using the prediction model, throughput increased by up to 4 aircraft per hour compared to a constant initial spacing.

A New View on Biodynamic Feedthrough Analysis: Unifying the Effects on Forces and Positions

When performing a manual control task, vehicle accelerations can cause involuntary limb motions, which can result in unintentional control inputs. This phenomenon is called biodynamic feedthrough (BDFT). In the past decades, many studies into BDFT have been performed, but its fundamentals are still only poorly understood. What has become clear, though, is that BDFT is a highly complex process, and its occurrence is influenced by many different factors. A particularly challenging topic in BDFT research is the role of the human operator, which is not only a very complex but also a highly adaptive system. In literature, two different ways of measuring and analyzing BDFT are reported. One considers the transfer of accelerations to involuntary forces applied to the control device (CD); the other considers the transfer of accelerations to involuntary CD deflections or positions. The goal of this paper is to describe an approach to unify these two methods. It will be shown how the results of the two methods relate and how this knowledge may aid in understanding BDFT better as a whole. The approach presented is based on the notion that BDFT dynamics can be described by the combination of two transfer dynamics: 1) the transfer dynamics from body accelerations to involuntary forces and 2) the transfer dynamics from forces to CD deflections. The approach was validated using experimental results.

Effects of Controlled Element Dynamics on Human Feedforward Behavior in Ramp-Tracking Tasks

In real-life manual control tasks, human controllers are often required to follow a visible and predictable reference signal, enabling them to use feedforward control actions in conjunction with feedback actions that compensate for errors. Little is known about human control behavior in these situations. This paper investigates how humans adapt their feedforward control dynamics to the controlled element dynamics in a combined ramp-tracking and disturbance-rejection task. A human-in-the-loop experiment is performed with a pursuit display and vehicle-like controlled elements, ranging from a single integrator through second-order systems with a break frequency at either 3, 2, or 1 rad/s, to a double integrator. Because the potential benefits of feedforward control increase with steeper ramp segments in the target signal, three steepness levels are tested to investigate their possible effect on feedforward control with the various controlled elements. Analyses with four novel models of the operator, fitted to time-domain data, reveal feedforward control for all tested controlled elements and both (nonzero) tested levels of ramp steepness. For the range of controlled element dynamics investigated, it is found that humans adapt to these dynamics in their feedforward response, with a close to perfect inversion of the controlled element dynamics. No significant effects of ramp steepness on the feedforward model parameters are found.