AeroSwarm: A GPS-Denied Multi-UAV Simulation Framework

AeroSwarm is a simulation framework designed to study collaborative drone navigation in GPS-denied environments. The platform models a swarm composed of one leader drone and two follower drones operating in a complex scenario with obstacles, limited visibility, and communication constraints.

https://video.wixstatic.com/video/300297_12c8844e8030442f81f102cc2d5b6580/720p/mp4/file.mp4

The leader drone is equipped with a 360° LiDAR sensor that continuously scans the environment, detects obstacles, and supports real-time navigation. During the mission, the leader transfers its position, waypoint information, velocity, LiDAR status, and navigation state to the follower drones through drone-to-drone communication links. This allows the swarm to maintain formation and continue coordinated navigation even when GPS is unavailable. The simulator includes multiple dashboard tabs for monitoring the mission. The Navigation + Mapping tab shows the camera views, HUD information, and swarm scenario map. The LiDAR + COMM overlay tab visualizes the LiDAR rays, obstacle detection, and communication links. The Sniffer tab monitors the packet-level communication between the leader and follower drones, including RSSI, latency, packet delivery ratio, RX/DROP state, and LOS/OCC channel condition. AeroSwarm is useful for testing autonomous navigation strategies, communication-aware formation control, LiDAR-based perception, and swarm coordination under GPS-denied conditions. It provides a flexible research platform for developing robust multi-UAV systems for search and rescue, inspection, defense, and operations in indoor, urban, forest, or degraded environments.

---

https://github.com/1Px-Vision/AeroSwarm

CSI-Channel Spatial Decomposition for WiFi-Based Human Pose Estimation in Rescue Scenarios

In rescue scenarios, locating and understanding human motion quickly can make the difference between a successful intervention and a delayed response. Yet cameras, LiDAR, and wearable sensors often struggle in smoke-filled rooms, collapsed buildings, dark environments, or areas where direct line of sight is blocked. This is where WiFi sensing offers a promising alternative. By analyzing how human movement alters wireless signal propagation, it becomes possible to estimate body posture without requiring a camera-based view of the person.

Human Identification Procedure

CSI is a measurement that represents these disparities between carriers and subcarriers. The signal received by the receiver can be expressed using the CSI as follows.

The proposed framework uses Channel State Information (CSI) as the main sensing source. Human presence and motion affect the wireless channel, creating measurable variations in signal amplitude and phase. These raw CSI signals are first collected and then passed through a structured preprocessing pipeline that improves signal quality and extracts the most relevant motion information. This includes antenna selection, outlier removal, discrete wavelet transform (DWT)-based processing, and feature extraction. The goal is to decompose the spatial and temporal characteristics of the WiFi channel into informative representations that can support robust pose estimation even in complex indoor environments.

Human Identification Procedure

It begins with CSI raw data acquisition, where both amplitude and phase are extracted from the WiFi signals. These measurements are then refined through data preprocessing and transformed into features suitable for learning. Finally, the neural network predicts the human skeletal configuration and reconstructs a 3D human body representation. This pipeline shows how wireless sensing can move beyond simple activity recognition toward detailed human pose recovery.

At the core of the method is a neural network architecture designed to reconstruct the human skeleton from the processed CSI data. The model combines an encoder-decoder structure with a Spatial Orientation Attention Module (SOA), allowing it to better capture body orientation and motion relationships across the wireless channel. This enables the system to infer a structured human pose representation and map it to a full-body model such as SMPL-X, providing a richer reconstruction of the person’s posture.

Workflow for WiFi-based human pose estimation using CSI data and neural networks.

On the left, CSI data collection combined with camera-based pose information is shown, where multiple wireless links capture signal variations caused by a human subject. In the center, the collected data is used for neural network design, which learns the relationship between CSI measurements and human body posture. The output of the network is a human skeleton reconstruction and SMPL-X body model, shown as both a skeletal representation and a 3D human figure. On the right, the figure highlights the dataset structure, including a sample 3×3 CSI matrix pose skeleton and the corresponding CSI amplitude signal, illustrating how wireless signal patterns are linked to body movement and posture.

Experimental Setup

The experimental setup and example outputs. It shows sample skeleton predictions alongside CSI amplitude signals captured across multiple wireless links. These examples demonstrate how different human poses, such as walking or crouching, produce distinguishable channel responses. The comparison between ground-truth and predicted skeletons provides visual evidence that CSI-based sensing can track meaningful posture changes with good spatial consistency. This approach is especially relevant for search-and-rescue operations, where first responders need situational awareness in environments that are unsafe or visually inaccessible. A WiFi-based human pose estimation system could help detect trapped survivors, infer whether a person is standing, lying down, or crouched, and support faster decision-making without intrusive sensing hardware. By leveraging existing wireless infrastructure, the method also opens the door to low-cost, privacy-preserving monitoring in emergency response, healthcare, and smart building applications.

Experimental examples show the relationship between CSI amplitude variations and predicted skeletal motion under different activities.

Takeaway

CSI-channel spatial decomposition shows that WiFi signals can be transformed into meaningful human pose information, enabling camera-free and privacy-preserving body reconstruction in difficult rescue environments. By combining CSI preprocessing, spatial feature learning, and lightweight neural modeling, this approach offers a practical path toward detecting and understanding human posture in smoke, darkness, debris, or non-line-of-sight conditions where conventional vision systems often fail.

---

LLM-Enabled Agent Stack for Denied Flight

Autonomous flight in GPS-denied environments is one of the most demanding challenges in modern robotics. Drones operating indoors, in forests, near collapsed infrastructure, inside industrial plants, or in dense urban canyons cannot depend on satellite navigation for stable localization and mission execution. In these scenarios, the aircraft must perceive the scene, interpret the mission, reason about obstacles, and generate safe actions using onboard intelligence.

A promising solution is the LLM-enabled agent stack for denied flight, where large language model reasoning is combined with perception, planning, control, and multi-agent coordination. Rather than using a single monolithic autonomy block, this architecture organizes the navigation system as a hierarchy of task-specific agents that cooperate through shared scene understanding and mission intent.

Why denied flight needs intelligent agent stacks

In denied environments, the drone faces several simultaneous difficulties:

• no continuous GPS position reference

• limited visibility caused by smoke, dust, shadows, or clutter

• narrow corridors and tight obstacle spacing

• uncertain or changing terrain structure

• dynamic mission goals and time pressure

Traditional navigation pipelines can handle parts of this problem, but they often struggle when the mission requires higher-level interpretation. A drone may detect obstacles and estimate motion, yet

still fail to decide the best route, adapt to new objectives, or coordinate with other drones. Where an LLM-enabled architecture becomes valuable. The language model acts as a reasoning layer that connects mission context, scene interpretation, route selection, and action priorities.

Architecture overview

The proposed architecture organizes denied-flight autonomy into a set of interacting task-specific modules. Each module contributes a dedicated function while exchanging information through structured scene and mission representations. The process begins with a perception layer, which extracts environmental cues from RGB images, depth information, and semantic masks. This layer identifies obstacles, free space, structures, and local scene features required for navigation.

The output of the perception layer is passed into a scene model, which constructs a compact representation of walls, hazards, traversable regions, and 3D geometry. This model provides a grounded description of the environment that can be used by both planning and reasoning modules. A prompt-driven task planning block introduces mission-level flexibility. Instead of relying only on hard-coded rules, the system can interpret high-level instructions such as corridor preference, inspection priority, formation behavior, or fallback conditions. This enables faster adaptation to new tasks and changing environments.

LLM agent navigation enables a drone to interpret mission context, analyze visual and state information, and make adaptive flight decisions in real time. By combining high-level reasoning with trajectory tracking and safety control, the agent can follow goals, avoid obstacles, and maintain stable navigation in dynamic or GPS-denied environments.

At the center of the architecture is the LLM reasoning layer. The LLM processes mission context, scene abstractions, and vehicle status to generate structured decisions that guide downstream modules. Its role is strategic rather than low-level. It may recommend path preferences, task priorities, coordination changes, or safety-aware responses when uncertainty increases. For team operations, a mission coordinator allocates roles and synchronizes multiple drones. This module supports cooperative exploration, target search, area partitioning, and mission reallocation. The LLM enhances this process by adding semantic mission awareness to geometric planning. The route selector and waypoint builder translate mission intent into executable subgoals. These modules select candidate paths and generate feasible intermediate waypoints that satisfy environmental and mission constraints.

A local guidance and safety layer then ensures physically safe execution. This block combines path-following control, obstacle avoidance, speed regulation, and action filtering. It acts as a safety shield between high-level reasoning and flight execution. Finally, a command aggregator fuses outputs from planning, coordination, and local control into a consistent kinematic command for the drone or drone group.

Functional significance

The main advantage of this stack lies in its separation of responsibilities. Perception grounds the system in real observations. The scene model organizes the environment into a usable structure. The LLM introduces contextual decision-making. Planning modules convert intent into motion objectives. Safety and control ensure that all generated commands remain feasible and secure. This layered design improves robustness in denied-flight conditions because the system is not dependent on a single navigation source. If localization becomes uncertain, mission logic can trigger fallback behaviors. If the environment changes, the planner can re-route. If multiple drones are available, coordination can be adjusted dynamically.

LLM-based multi-UAV mission planning system that converts user commands and external intelligence into coordinated flight tasks and formation control.

The LLM-enabled agent stack for denied flight provides a modular and intelligent framework for autonomous aerial navigation in challenging environments. By combining perception, scene understanding, prompt-driven planning, mission coordination, and safety-constrained control, the architecture supports adaptive and cooperative drone behavior when GPS is not available. This approach represents a strong direction for next-generation aerial autonomy, especially in missions that require resilience, flexibility, and higher-level reasoning.