Abstract:We propose a new approach to visual perception for relative localization of agents within large-scale swarms of UAVs. Inspired by biological perception utilized by schools of sardines, swarms of bees, and other large groups of animals capable of moving in a decentralized yet coherent manner, our method does not rely on detecting individual neighbors by each agent and estimating their relative position, but rather we propose to regress a neighbor density over distance. This allows for a more accurate distance estimation as well as better scalability with respect to the number of neighbors. Additionally, a novel swarm control algorithm is proposed to make it compatible with the new relative localization method. We provide a thorough evaluation of the presented methods and demonstrate that the regressing approach to distance estimation is more robust to varying relative pose of the targets and that it is suitable to be used as the main source of relative localization for swarm stabilization.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to provide a new visual perception method for large - scale unmanned aerial vehicle (UAV) swarms to achieve relative positioning. Specifically, the paper proposes a biologically - inspired visual relative positioning method, aiming to solve the following problems: 1. **Relative positioning of large - scale UAV swarms**: Traditional detection - based methods have problems of poor scalability and excessive complexity when dealing with a large number of targets. The paper proposes a method to estimate distance by regressing neighbor density instead of relying on each UAV to detect and estimate the relative position of its neighbors. 2. **Improving the accuracy of distance estimation**: Traditional methods are easily affected by the target pose change when estimating the relative position of the target, resulting in estimation deviation. The new method improves the accuracy and robustness of distance estimation by regressing the UAV density within the distance range. 3. **Control algorithm adapted to agile group behavior**: In order to make the newly proposed relative positioning method match the behavior of the UAV swarm, the paper also proposes a new swarm control algorithm to ensure that the UAV swarm can avoid collisions while maintaining stability. ### Formula summary - **Neighbor density distribution**: \[ l_{\text{raw}}[d]=\left|\{p \mid p \in M,\|p\| \in [d, d + \Delta d]\}\right| \] where \(d\) is the distance from the camera to the target, \(\Delta d\) is the distance discretization step size, \(M\) is the set of UAV positions in the image grid cell, and \(p\) is the position of the UAV (represented in the camera optical coordinate system). - **Partially smoothed ground - truth label**: \[ l_{\text{gt}}[d]=(l_{\text{raw}}*g)[d] \] where \(g[d]\) is defined as: \[ g[d]= \begin{cases} \delta[d], & \text{for the nearest } k \text{ bins}\\ G_\sigma[d], & \text{otherwise} \end{cases} \] Here \(G_\sigma[d]\) represents a Gaussian kernel with a standard deviation of \(\sigma\), \(\delta[d]\) is the unit impulse function, and \((f * g)[x]\) represents the discrete convolution of functions \(f[x]\) and \(g[x]\) with respect to \(x\). - **Loss function**: \[ L = \|(l_o - l_{\text{gt}})\odot w\| \] where \(l_{\text{gt}}\) is the corresponding ground - truth density vector, \(w\) is a non - learnable distance bin weight vector, and the symbol \(\odot\) represents the element - wise multiplication between vector elements. ### Conclusion The new method proposed in the paper solves the scalability and accuracy problems in the relative positioning of large - scale UAV swarms by regressing the density of UAVs in different distance ranges, and verifies its effectiveness in swarm control and collision avoidance through experiments.

Bio-inspired visual relative localization for large swarms of UAVs

Cost-effective Swarm Navigation System Via Close Cooperation

Relative Localization within a Quadcopter Unmanned Aerial Vehicle Swarm Based on Airborne Monocular Vision

Bio-Inspired Compact Swarms of Unmanned Aerial Vehicles without Communication and External Localization

Agile Formation Control of Drone Flocking Enhanced With Active Vision-Based Relative Localization

Vision-based Drone Flocking in Outdoor Environments

Distributed surveillance by a swarm of UAVs operating under positional uncertainty

Fusion of Visual-Inertial Odometry with LiDAR Relative Localization for Cooperative Guidance of a Micro-Scale Aerial Vehicle

Monocular Vision based Collaborative Localization for Micro Aerial Vehicle Swarms

Fast and Omnidirectional Relative Position Estimation With Circular Markers for UAV Swarm

Onboard Ranging-based Relative Localization and Stability for Lightweight Aerial Swarms

SBC-SLAM: Semantic Bioinspired Collaborative SLAM for Large-Scale Environment Perception of Heterogeneous Systems

Distributed and Collaborative Localization for Swarming UAVs

Learning Vision-Based Flight in Drone Swarms by Imitation

Distributed 3D Relative Localization of UAVs

Infrastructure-free Localization of Aerial Robots with Ultrawideband Sensors

Fast Collective Evasion in Self-Localized Swarms of Unmanned Aerial Vehicles

Tag-based Visual Odometry Estimation for Indoor UAVs Localization

Preserving Relative Localization of FoV-Limited Drone Swarm via Active Mutual Observation

Resonant Beam Enabled Relative Localization for UAV Swarm

System for deployment of groups of unmanned micro aerial vehicles in GPS-denied environments using onboard visual relative localization