VG-Swarm: A Vision-based Gene Regulation Network for UAVs Swarm Behavior Emergence

TL;DR

VG-Swarm is presented, a practical and effective method for aerial robots dynamic encirclement, which consists of a vision-based gene regulatory network (V-GRN) and a visual perception module that can emerge adaptive pattern formations to entrap the targets even without any communication and global information.

Abstract

Unmanned Aerial Vehicles (UAVs) dynamic encirclement is an emerging field with great potential. Researchers often get inspiration from biological systems, either from macro-world like fish schools or bird flocks etc, or from micro-world like gene regulatory networks (GRN). However, most swarm control algorithms rely on centralized control, global information acquisition, and communications among neighboring agents. In this work, we propose a distributed swarm control method based purely on vision and GRN without any direct communications, in which swarm agents of e.g. UAVs can generate an entrapping pattern to encircle an escaping target of UAV based purely on their installed omnidirectional vision sensors. A finite-state-machine (FSM) describing the behavioral model of each drone is also designed so that a swarm of drones can accomplish searching and entrapping of the target collectively in an integrated way. We verify the effectiveness and efficiency of the proposed method in various simulation and real-world experiments.

Figures (14)

• Figure 1: The photo of target entrapping pattern formation for UAVs based on V-GRN in a real world experiment. Each UAV detects and estimates the positions of targets, obstacles and neighbors through omnidirectional images, taking the UAV at the bottom of the photo as an example.
• Figure 2: The system architecture of the proposed VG-Swarm. The input, omnidirectional images captured from four cameras, is used for object detection and position estimation. In the central part, a local map records the position information of the observed objects, and V-GRN generates target entrapping patterns using the environmental information provided by the local map. Then the planner generates the command to control the agent's movement according to the FSM model and the calculated data from V-GRN.
• Figure 3: The schematic diagram illustrating the local coordinate system, generation of entrapping patterns and its corresponding experimental implementation. The UAV establishes a right-handed coordinate system with the direction of its own nose as the y-axis. Part B illustrate the entrapping pattern formed by four local patterns. In part C, the final entrapping pattern of the four UAVs is formed in the experiment. Part D shows the visual perception of the UAV in the lower right corner of part C.
• Figure 4: A diagram of the vision-based gene regulatory network (V-GRN). A cell represents an individual UAV, which consists of an upper layer and a lower layer. In the upper layer, $p_1$, $p_2$, and $p_3$ are sensory proteins that can be regulated by external environmental inputs from the targets, obstacles, and neighbors, respectively. $M$ is a protein that fuses the concentration field from proteins $p_1$, $p_2$, $p_3$, and affects the production of proteins $G_1$ and $G_2$, which are actuating proteins in the bottom layer representing control output of orientation and velocity value, respectively. They both affect the production of protein $P$, which ultimately determines the dynamic position of the UAV. In a system, protein $P$ of one cell affects the gene expression of other neighboring cells.
• Figure 5: The concentration fields diagram. (a) is the concentration field formed by a target, (b) is the concentration field formed by a obstacle or neighboring UAV, and (c) is the comprehensive concentration field.