.png){kind=icon}
.png){kind=icon}
Return
Bayesian Optimization Based Trust Model for Human Multi-robot Collaborative Motion Tasks in Offroad Environments
.png){kind=icon}
PRE
AI
DOI：10.1007/s12369-023-01011-2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this paper, we seek to develop a computational human to multi-robot system (MRS) trust model to encode human intention into the MRS motion tasks in offroad environments. Our computational trust model builds a linear state-space equation to capture the influence of environmental attributes on human trust in an MRS. Bayesian inference is used to derive the posterior distribution of the trust model parameters. Due to the intractable computation of the posterior distributions, we develop a Markov Chain Monte Carlo sampling algorithm by integrating the Gibbs sampler with the forward-filtering-backward-sampling to approximate the distributions. A Bayesian optimization based experimental design (BOED) is proposed to sequentially learn the human-MRS trust model parameters. Inspired by decision field theory, we develop a human preference based acquisition function for the BOED to explore the MRS motion path and collect data for the trust model in an efficient way. A case study on human-MRS collaborative bounding overwatch task is deployed, which is a multi-robot motion task traditionally used in offroad environments and requires a heavy cognition workload for the human to collaborate with the MRS. Trials using simulated human agents and human subjects collaborating with an MRS are conducted in the ROS Gazebo simulator. The simulated human agent with MRS shows that the BOED can correctly estimate the trust model parameters. The human subject tests demonstrate the capability of our computational trust model in capturing the human's trust dynamics with the goodness of fit metrics. The tests also show statistically significant results by comparing the BOED with a benchmark experimental design approach. The BOED resulted in fewer collisions with obstacles, lower frequency of contact loss between robots, lower operator workload, and higher system usability.
Keywords:
Computational trust models
Multi-robot systems
Bayesian optimization
Bounding overwatch
Journal
IF:
3.7
Papers:
1.4K
Citations:
5.6K