Abstract:Human behavior modeling is important for the design and implementation of human-automation interactive control systems. In this context, human behavior refers to a human's control input to systems. We propose a novel method for human behavior modeling that uses human demonstrations for a given task to infer the unknown task objective and the variability. The task objective represents the human's intent or desire. It can be inferred by the inverse optimal control and improve the understanding of human behavior by providing an explainable objective function behind the given human behavior. Meanwhile, the variability denotes the intrinsic uncertainty in human behavior. It can be described by a Gaussian mixture model and capture the uncertainty in human behavior which cannot be encoded by the task objective. The proposed method can improve the prediction accuracy of human behavior by leveraging both task objective and variability. The proposed method is demonstrated through human-subject experiments using an illustrative quadrotor remote control example.

What problem does this paper attempt to address?

This paper mainly discusses how to model human behavior by identifying task goals and behavioral changes, which is important for designing and implementing human-computer interaction control systems. The authors propose a new approach that utilizes human demonstrations of a given task to infer unknown task goals and changes. Task goals reflect people's intentions or desires, and can be inferred through inverse optimal control; while changes describe the inherent uncertainty of human behavior and can be represented using Gaussian mixture models. The paper points out that existing modeling techniques either rely on pre-defined structures or can only replicate result trajectories without explaining the underlying reasons for behavior. Therefore, their proposed method combines both task goals and changes to improve the accuracy of behavior prediction. Through experiments, this method is validated in a quadcopter remote control example, demonstrating its advantages in explaining behavior, improving prediction accuracy, and data efficiency. Specifically, the paper first introduces the background of the problem and human behavior models, and then proposes a model that combines task goals and changes to represent behavior. The task goal model is based on inverse optimal control, while the change model adopts Gaussian mixture models. Next, the paper elaborates on the parameter identification methods, including inverse optimal control and Gaussian mixture model parameter estimation. Finally, the effectiveness of this method is demonstrated through human subject experiments and compared with existing methods. In conclusion, this paper aims to address how to accurately understand and predict human behavior, particularly considering the goal-orientedness and uncertainty in behavior, which is highly significant for the design and performance improvement of human-computer interaction systems.

Human Behavior Modeling via Identification of Task Objective and Variability

Human-Variability-Respecting Optimal Control for Physical Human-Machine Interaction

Study on Human-Variability-Respecting Optimal Control Affecting Human Interaction Experience

Models of human behavior for human-robot interaction and automated driving: How accurate do the models of human behavior need to be?

Towards Modeling and Influencing the Dynamics of Human Learning

A Universal Self-Adaption Workspace Mapping Method for Human–Robot Interaction Using Kinect Sensor Data

Quantifying Hypothesis Space Misspecification in Learning from Human-Robot Demonstrations and Physical Corrections

Modeling Human Behavior Part I -- Learning and Belief Approaches

Performance Evaluation and Optimization of Human Control Strategy.

Inferring control objectives in a virtual balancing task in humans and monkeys

Robust Robot Planning for Human-Robot Collaboration

Human stochastic closed-loop behavior for master-slave teleoperation using multi-leap-motion sensor

Adaptive Intention-Driven Variable Impedance Control for Wearable Robots with Compliant Actuators

Modeling sensory-motor decisions in natural behavior

Human Behavior Simulation: Objectives, Methodologies, and Open Problems

On the Sensitivity of Reward Inference to Misspecified Human Models

Modeling Human Behavior Part II -- Cognitive approaches and Uncertainty

Generative Modeling of Multimodal Multi-Human Behavior

Biodynamic modeling, system identification, and variability of multi-finger movements.

A Motion Control Approach for Physical Human–Robot-Environment Interaction Via Operational Behaviors Inference

Probabilistic Multimodal Modeling for Human-Robot Interaction Tasks