# Cooperative Sensing Enhanced UAV Path-Following and Obstacle Avoidance with Variable Formation

## TL;DR

This paper presents a comprehensive approach combining deep reinforcement learning, integrated sensing, and hierarchical fusion to improve UAV path-following and obstacle avoidance, ensuring safe and efficient flight in complex environments.

## Contribution

It introduces a novel DRL-based formation path-following model, an ISAC-based obstacle sensing method with CRLB analysis, and a hierarchical fusion strategy for conflict-free subtask integration.

## Key findings

- The proposed DRL model achieves high accuracy in path-following.
- The ISAC-based sensing with CRLB provides precise obstacle localization.
- Hierarchical fusion improves overall UAV navigation safety and efficiency.

## Abstract

The high mobility of unmanned aerial vehicles (UAVs) enables them to be used in various civilian fields, such as rescue and cargo transport. Path-following is a crucial way to perform these tasks while sensing and collision avoidance are essential for safe flight. In this paper, we investigate how to efficiently and accurately achieve path-following, obstacle sensing and avoidance subtasks, as well as their conflict-free fusion scheduling. Firstly, a high precision deep reinforcement learning (DRL)-based UAV formation path-following model is developed, and the reward function with adaptive weights is designed from the perspective of distance and velocity errors. Then, we use integrated sensing and communication (ISAC) signals to detect the obstacle and derive the Cramer-Rao lower bound (CRLB) for obstacle sensing by information-level fusion, based on which we propose the variable formation enhanced obstacle position estimation (VFEO) algorithm. In addition, an online obstacle avoidance scheme without pretraining is designed to solve the sparse reward. Finally, with the aid of null space based (NSB) behavioral method, we present a hierarchical subtasks fusion strategy. Simulation results demonstrate the effectiveness and superiority of the subtask algorithms and the hierarchical fusion strategy.