Frank M. Drop

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.

Identification of the transition from compensatory to feedforward behavior in manual control

The human in manual control of a dynamical system can use both feedback and feedforward control strategies and will select a strategy based on performance and required effort. Literature has shown that feedforward control is used during tracking tasks in response to predictable targets. The influence of an external disturbance signal on the utilization of a feedforward control strategy has never been investigated, however. We hypothesized that the human will use a combined feedforward and feedback control strategy whenever the predictable target signal is sufficiently strong, and a predominantly feedback strategy whenever the random disturbance signal is dominant. From the data of a human-in-the-loop experiment we conclude that feedforward control is used in all the considered experimental conditions, including those where the disturbance signal is dominant and feedforward control does not deliver a marked performance advantage.

Objective Model Selection for Identifying the Human Feedforward Response in Manual Control

Realistic manual control tasks typically involve predictable target signals and random disturbances. The human controller (HC) is hypothesized to use a feedforward control strategy for target-following, in addition to feedback control for disturbance-rejection. Little is known about human feedforward control, partly because common system identification methods have difficulty in identifying whether, and (if so) how, the HC applies a feedforward strategy. In this paper, an identification procedure is presented that aims at an objective model selection for identifying the human feedforward response, using linear time-invariant autoregressive with exogenous input models. A new model selection criterion is proposed to decide on the model order (number of parameters) and the presence of feedforward in addition to feedback. For a range of typical control tasks, it is shown by means of Monte Carlo computer simulations that the classical Bayesian information criterion (BIC) leads to selecting models that contain a feedforward path from data generated by a pure feedback model: “false-positive” feedforward detection. To eliminate these false-positives, the modified BIC includes an additional penalty on model complexity. The appropriate weighting is found through computer simulations with a hypothesized HC model prior to performing a tracking experiment. Experimental human-in-the-loop data will be considered in future work. With appropriate weighting, the method correctly identifies the HC dynamics in a wide range of control tasks, without false-positive results.

Effects of Target Signal Shape and System Dynamics on Feedforward in Manual Control

The human controller (HC) in manual control of a dynamical system often follows a visible and predictable reference path (target). The HC can adopt a control strategy combining closed-loop feedback and an open-loop feedforward response. The effects of the target signal waveform shape and the system dynamics on the human feedforward dynamics are still largely unknown, even for common, stable, vehicle-like dynamics. This paper studies the feedforward dynamics through computer model simulations and compares these to system identification results from human-in-the-loop experimental data. Two target waveform shapes are considered, constant velocity ramp segments and constant acceleration parabola segments. Furthermore, three representative vehicle-like system dynamics are considered: 1) a single integrator (SI); 2) a second-order system; and 3) a double integrator. The analyses show that the HC utilizes a combined feedforward/feedback control strategy for all dynamics with the parabola target, and for the SI and second-order system with the ramp target. The feedforward model parameters are, however, very different between the two target waveform shapes, illustrating the adaptability of the HC to task variables. Moreover, strong evidence of anticipatory control behavior in the HC is found for the parabola target signal. The HC anticipates the future course of the parabola target signal given extensive practice, reflected by negative feedforward time delay estimates.

Control-Theoretic Models of Feedforward in Manual Control

Understanding how humans control a vehicle (cars, aircraft, bicycles, etc.) enables engineers to design faster, safer, more comfortable, more energy efficient, more versatile, and thus better vehicles. In a typical control task, the Human Controller (HC) gives control inputs to a vehicle such that it follows a particular reference path (e.g., the road) accurately. The HC is simultaneously required to attenuate the effect of disturbances (e.g., turbulence) perturbing the intended path of the vehicle. To do so, the HC can use a control organization that resembles a closed-loop feedback controller, a feedforward controller, or a combination of both. Previous research has shown that a purely closed-loop feedback control organization is observed only in specific control tasks, that do not resemble realistic control tasks, in which the information presented to the human is very limited. In realistic tasks, a feedforward control strategy is to be expected; yet, almost all previously available HC models describe the human as a pure feedback controller lacking the important feedforward response. Therefore, the goal of the research described in this thesis was to obtain a fundamental understanding of feedforward in human manual control. First, a novel system identification method was developed, which was necessary to identify human control dynamics in control tasks involving realistic reference signals. Second, the novel identification method was used to investigate three important aspects of feedforward through human-in-the-loop experiments which resulted in a control-theoretical model of feedforward in manual control. The central element of the feedforward model is the inverse of the vehicle dynamics, equal to the theoretically ideal feedforward dynamics. However, it was also found that the HC is not able to apply a feedforward response with these ideal dynamics, and that limitations in the perception, cognition, and action loop need to be modeled by additional model elements: a gain, a time delay, and a low-pass filter. Overall, the thesis demonstrated that feedforward is indeed an essential part of human manual control behavior and should be accounted for in many human-machine applications.

Objective ARX Model Order Selection for Multi-Channel Human Operator Identification

In manual control, the human operator primarily responds to visual inputs but may elect to make use of other available feedback paths such as physical motion, adopting a multi-channel control strategy. Hu- man operator identification procedures generally require a priori selection of the model structure, which can be problematic as the exact feedback organization operators adopt is not always clear in advance. This pa- per evaluates a novel method for objectively detecting the presence of additional human operator feedback responses in control tasks with multiple inputs. The approach makes use of linear-time invariant ARX mod- els for system identification, combined with an objective model selection criterion. To test the method, an experiment was conducted in which participants performed a compensatory yaw attitude tracking task in a moving-base flight simulator, with varying motion cueing settings. In addition, a pursuit tracking condition without motion feedback was tested. For all conditions, the objective ARX model-based identification method was used to verify the presence of a possible additional human operator output feedback response. With ap- propriate tuning of the penalty on model complexity in the model selection criterion, the methodology proved successful in correctly identifying the additional operator responses in experimental conditions that contained no motion or high-quality motion feedback. With low-fidelity motion feedback or a pursuit display, the results suggest that no consistent feedback response is achieved by the participants. The approach was substantiated with offline Monte Carlo simulations, which show strong correlation with the obtained experiment results.