ISAC-Enabled Multi-UAV Collaborative Target Sensing for Low-Altitude Economy

This paper proposes a low-complexity iterative algorithm for an ISAC-enabled multi-UAV collaborative target sensing scheme that jointly optimizes UAV-BS association, trajectories, and bandwidth allocation to minimize the posterior Cramer-Rao bound for tracking dynamic low-altitude targets while satisfying communication constraints.

The Gist

Imagine a busy city sky filled with delivery drones, air taxis, and surveillance bots. This is the "Low-Altitude Economy." While these machines are great for delivering pizza or checking power lines, there's a problem: what if an unauthorized drone (a "rogue" drone) flies in and threatens safety? Or what if a delivery drone gets lost?

We need a way to see (sense) these targets clearly and talk (communicate) to them instantly. But fixed ground towers (Base Stations) are like lighthouses; they can't move. If a target hides behind a building or moves too fast, the lighthouse might lose sight of it.

This paper proposes a solution using swarms of friendly drones that act like a team of "super-scouts." Here is the breakdown of how it works, using simple analogies:

1. The Super-Tool: ISAC (The Swiss Army Knife)

Traditionally, a drone uses one radio to talk to the ground and a separate radar to look for things. This is like having a walkie-talkie in one hand and a flashlight in the other.
ISAC (Integrated Sensing and Communication) is like a Swiss Army Knife. It combines the flashlight and the walkie-talkie into one tool. The drone sends out a signal that does two things at once:

• It talks to the ground station to send data.
• It bounces off the target (like an echo) to tell the drone exactly where the target is and how fast it's moving.

2. The Challenge: The Moving Target

The problem is that the "rogue" target is unpredictable. It's like trying to catch a slippery fish in a dark river.

• Fixed towers can't move to get a better view.
• The drones need to fly to the right spot to see the fish, but they also need to stay close enough to the towers to send their data back without losing the connection.
• The Math: If the drones are too far apart or in a straight line with the target, they get a blurry picture. They need to form a specific geometric shape (like a triangle) around the target to get a sharp, 3D view.

3. The Solution: A Smart Dance

The authors propose a system where multiple drones work together in a dynamic dance. They don't just fly randomly; they calculate the perfect moves in real-time.

The system optimizes three things simultaneously:

1. Who talks to whom? (Which drone connects to which ground tower?)
2. Where do they fly? (The flight path).
3. How much "airtime" do they use? (Bandwidth allocation).

Think of it like a conductor leading an orchestra:

• The Conductor (the algorithm) tells the Violins (some drones) to fly closer to the target to get a better "sound" (sensing).
• It tells the Cellos (other drones) to stay a bit further back to ensure they have a strong connection to the ground (communication).
• It decides how much of the "stage time" (bandwidth) each instrument gets so the music (data) doesn't get garbled.

4. The "Crystal Ball" (PCRB)

How does the system know if it's doing a good job? It uses a mathematical concept called PCRB (Posterior Cramér-Rao Bound).

• Simple Analogy: Imagine you are guessing where a ball is rolling in the dark. The PCRB is like a "Guaranteed Worst-Case Error." It tells you, "Even with the best guess possible, your error will be at least this big."
• The goal of the paper is to minimize this error. The algorithm constantly asks, "If I move this drone 10 meters left and give it more bandwidth, will my 'guaranteed error' get smaller?" If yes, it moves the drone.

5. The Algorithm: The "Step-by-Step" Dancer

Solving all these math problems at once is incredibly hard (like trying to solve a Rubik's cube while juggling). The authors created a low-complexity iterative algorithm.

• Step 1: First, make sure the drones can actually talk to the ground. If they can't, the mission fails.
• Step 2: Once they are connected, the algorithm starts "dancing." It moves the drones a little bit, checks if the view got better, then moves them again. It keeps adjusting the flight path and the data bandwidth until the "blur" on the target is as small as possible.

6. The Results: Seeing Clearly

When they tested this in a computer simulation:

• Static Drones (sitting still) were like trying to take a photo of a race car from a distance; the picture was blurry.
• The Proposed Method (the smart dance) allowed the drones to fly right next to the target and surround it.
• The Outcome: The system reduced the error in guessing the target's location by over 60% compared to the old ways. It was much faster and more accurate at spotting the "rogue" drone.

Summary

In short, this paper teaches us how to turn a swarm of drones into a smart, moving radar net. Instead of waiting for a target to appear in a fixed camera's view, the drones actively chase the target, adjust their formation like a flock of birds, and share their data efficiently to ensure we never lose sight of what's happening in the low-altitude sky. This is a crucial step toward making our future skies safe and efficient.

Technical Summary

Here is a detailed technical summary of the paper "ISAC-Enabled Multi-UAV Collaborative Target Sensing for Low-Altitude Economy."

1. Problem Statement

The paper addresses the challenge of sensing dynamic, unauthorized targets (e.g., illegal drones) within the Low-Altitude Economy (LAE) using Integrated Sensing and Communication (ISAC) technology.

• Context: While fixed ground Base Stations (BSs) provide ISAC services, they struggle with the high dynamics of low-altitude targets, particularly in coverage blind zones or at the edge of coverage.
• Core Challenge: Designing a collaborative system where multiple Unmanned Aerial Vehicles (UAVs) must simultaneously:
  1. Communicate with terrestrial BSs to maintain reliable control links.
  2. Sense a mobile target with unknown and dynamic location/velocity.
  3. Optimize their flight trajectories and resource allocation (bandwidth) in real-time.
• Specific Difficulties: The problem involves a non-convex and implicit objective function with coupled variables (UAV-BS association, trajectories, and bandwidth). The target's state is unknown, requiring real-time updates based on predictions, making pre-designing trajectories impossible.

2. Methodology

The authors propose a Dynamic Collaborative Target Sensing Scheme driven by a Posterior Cramér-Rao Bound (PCRB) minimization framework.

A. System Model

• Scenario: $M$ UAVs, $K$ BSs, and 1 dynamic target. UAVs reuse ISAC signals for both communication (to BSs) and sensing (target echo).
• Communication Model: Uses Line-of-Sight (LoS) channels with Frequency Division Multiple Access (FDMA). Constraints include minimum transmission rates ($R_{th}$) and SNR thresholds for reliable control.
• Sensing Model:
  • Target Evolution: Modeled as a nearly constant velocity process with Gaussian process noise.
  • Measurements: UAVs obtain range and velocity from echo signals. Measurement noise variance is inversely proportional to the Signal-to-Noise Ratio (SNR) of the echo, which depends on distance and allocated bandwidth.
  • Performance Metric: The PCRB is derived as the lower bound for the estimation error covariance of the target state (location and velocity). Minimizing PCRB maximizes sensing accuracy.
• Mechanism: A closed-loop system using Extended Kalman Filtering (EKF):
  1. Predict target state.
  2. Optimize UAV trajectories and bandwidth to minimize predicted PCRB.
  3. Execute sensing/communication.
  4. Update target state.

B. Optimization Problem

The goal is to minimize the trace of the PCRB matrix ($tr(C)$) by jointly optimizing:

• UAV-BS Association ($A$): Which BS each UAV connects to.
• UAV Trajectories ($Q$): Horizontal positions over time.
• Bandwidth Allocation ($\eta$): Fraction of total bandwidth assigned to each UAV.

Constraints:

• Minimum communication rate and SNR for UAV control.
• Maximum UAV speed and collision avoidance (safe distance between UAVs and between UAVs and the target).
• BS capacity limits.

C. Proposed Algorithm

Due to the non-convex nature of the problem, the authors propose a low-complexity iterative algorithm based on an Alternating Optimization (AO) framework:

1. Feasibility Check & Association:
   • First, determine the optimal UAV-BS association by solving a communication-maximization subproblem.
   • UAVs associate with the nearest BS that satisfies communication constraints to maximize the communication rate.
2. Alternating Optimization (AO):
   • Trajectory Optimization: With fixed bandwidth and association, optimize UAV paths. Since the objective is non-convex, a Descent Direction Search Algorithm is used. It approximates the objective via first-order Taylor expansion and uses Successive Convex Approximation (SCA) to handle non-convex constraints (collision avoidance, communication range).
   • Bandwidth Allocation: With fixed trajectories, optimize bandwidth distribution. This subproblem is convex and solved directly.
3. Iteration: The process repeats until convergence (change in objective value falls below a threshold $\epsilon$).

3. Key Contributions

1. Novel Scheme: Proposed an ISAC-enabled multi-UAV dynamic collaborative sensing scheme specifically for the LAE, allowing UAVs to dynamically adjust flight paths and resources for mobile target tracking.
2. Theoretical Derivation: Derived the PCRB for the target state as the sensing metric, explicitly linking sensing error to UAV geometry and bandwidth allocation.
3. Algorithm Design: Developed a low-complexity iterative algorithm that decouples the complex joint optimization problem into manageable subproblems (Association, Trajectory, Bandwidth) using AO and descent direction search.
4. Performance Validation: Demonstrated through simulations that the joint optimization significantly outperforms static deployment and average allocation strategies.

4. Simulation Results

The authors validated the scheme using MATLAB with a 3km $\times$ 3km area, 5 BSs, and 3 UAVs tracking a target.

• Trajectory Behavior: UAVs initially fly rapidly toward the target to reduce distance (improving SNR). As they get closer, they adjust their azimuth angles to create a more orthogonal geometric distribution relative to the target, which improves localization accuracy.
• Bandwidth Allocation: Resources are dynamically allocated to UAVs with better geometric positions to maximize sensing gains.
• Performance Comparison:
  • Compared against Static Deployment (SD), Average Bandwidth Allocation (ABA), and a CRB-criterion algorithm.
  • PCRB Reduction: The proposed method achieved the lowest PCRB throughout the mission.
  • Error Reduction:
    • Location Error: Reduced by 61.1% (vs. SD), 22.8% (vs. ABA), and 9.0% (vs. CRB-criterion).
    • Velocity Error: Reduced by 59.9%, 29.1%, and 12.0% respectively.

5. Significance

• Enabling LAE Safety: This work provides a critical technical foundation for ensuring the safety of the Low-Altitude Economy by enabling real-time, high-precision tracking of unauthorized intruders in complex urban environments.
• Resource Efficiency: It demonstrates that ISAC can effectively share spectrum and hardware resources between communication and sensing, maximizing the utility of 6G networks.
• Dynamic Adaptability: Unlike fixed systems, the proposed UAV-based approach adapts to target mobility, overcoming coverage blind spots and maintaining high sensing quality even for fast-moving targets.
• Scalability: The proposed algorithm has polynomial complexity ($O(M^3)$), making it feasible for real-time implementation with a moderate number of UAVs.

In conclusion, the paper presents a robust framework for next-generation aerial surveillance, proving that joint optimization of trajectory and communication resources is essential for high-precision dynamic target sensing in 6G-enabled low-altitude networks.