HI-GVF: Shared Control based on Human-Influenced Guiding Vector Fields for Human-multi-robot Cooperation

TL;DR

The paper tackles the burden and rigidity of conventional leader–follower multi-robot systems by introducing HI-GVF, a layered shared-control framework that lets a human specify a desired path and propagates this intention through a robot-formation via an intention field. A upper-layer intention-field model blends human input with robot and neighbor intentions, while a lower-layer policy-blending network fuses the resulting shared intention with the formation VF to achieve robust coordination. The approach lever obstacle-aware HI-GVF with reactive and repulsive boundaries and smooth bump functions, and enforces safety using safety barrier certificates with centralized and distributed QP formulations. Validation includes simulations and physical experiments across BCI, EMG, and eye-tracking interfaces in firefighting-like tasks, demonstrating improved target prioritization, faster response to occlusions, and reliable obstacle avoidance. The work provides formal stability analyses for both the intentional field and the consensus network, supporting the practical deployment of HI-GVF for scalable, safe human–multi-robot collaboration.

Abstract

Human-multi-robot shared control leverages human decision-making and robotic autonomy to enhance human-robot collaboration. While widely studied, existing systems often adopt a leader-follower model, limiting robot autonomy to some extent. Besides, a human is required to directly participate in the motion control of robots through teleoperation, which significantly burdens the operator. To alleviate these two issues, we propose a layered shared control computing framework using human-influenced guiding vector fields (HI-GVF) for human-robot collaboration. HI-GVF guides the multi-robot system along a desired path specified by the human. Then, an intention field is designed to merge the human and robot intentions, accelerating the propagation of the human intention within the multi-robot system. Moreover, we give the stability analysis of the proposed model and use collision avoidance based on safety barrier certificates to fine-tune the velocity. Eventually, considering the firefighting task as an example scenario, we conduct simulations and experiments using multiple human-robot interfaces (brain-computer interface, myoelectric wristband, eye-tracking), and the results demonstrate that our proposed approach boosts the effectiveness and performance of the task.

Figures (16)

• Figure 1: Human and multi-robot collaboration.
• Figure 2: The framework of the proposed human-influenced guiding vector field shared control method. The human operator and each robot are independent units, ultimately acting on the MRS through shared control. Different colored blocks are used to distinguish the different research contents in this figure.
• Figure 3: The visualization of a guiding vector field. The corresponding desired path (red curve) is the circle described by choosing $\phi(x,y) = x^2 + y^2 - R^2$, where $R$ is the circle radius, in (1). Each blue arrow represents a vector of the corresponding vector field at the position
• Figure 4: Illustration of the reactive and repulsive boundary and area of one obstacle.
• Figure 5: Illustration of smooth zero-out and zero-in bump functions.

Theorems & Definitions (8)

• Theorem 1
• proof
• Lemma 1
• Lemma 2
• proof
• Theorem 2
• proof
• Corollary 1