In this work, we present a comprehensive study on cooperative formation control algorithms and collision avoidance strategies for fixed-wing drones. Focusing on a leader-follower architecture, we develop a flight dynamics model tailored for fixed-wing drones, design separate control laws for the leader and followers, and introduce a distributed communication topology to enhance information exchange. A collision avoidance mechanism is integrated to prevent inter-drone collisions. Simulation results validate the effectiveness of the proposed methods in maintaining formation and avoiding collisions. Throughout the paper, we emphasize the unique challenges and solutions associated with fixed-wing drones, such as nonholonomic constraints and velocity limitations.
1. Introduction
The use of fixed-wing drones has expanded rapidly in military and civilian applications due to their long endurance, high speed, and large payload capacity. However, single fixed-wing drones suffer from limited operational range and poor robustness against disturbances. Multi-drone formations composed of fixed-wing drones can overcome these limitations by enabling cooperative mission execution. Formation control is a fundamental technology for multi-drone systems, requiring precise coordination of positions, velocities, and attitudes. Among various control approaches, the leader-follower method is intuitive and easy to implement, but it often relies on centralized communication which may cause single-point failures. To address this, we adopt a distributed topology that allows neighboring fixed-wing drones to exchange state information, improving fault tolerance and scalability.
In this paper, we propose a formation control algorithm for fixed-wing drones based on a leader-follower structure combined with distributed communication. The leader tracks a predefined trajectory, while followers adjust their velocities and roll angles to follow the leader and maintain the desired formation geometry. Additionally, we incorporate a collision avoidance strategy that dynamically switches between normal formation control and a temporary avoidance mode when the risk of collision is detected. Through extensive simulations, we demonstrate the feasibility and robustness of our approach for fixed-wing drones operating under realistic constraints.
2. Formation Model
2.1 Dynamics of Fixed-Wing Drones
We consider a fixed-wing drone as a nonholonomic nonlinear system. Under the assumptions of level flight (small pitch angle) and negligible vertical acceleration, the kinematic equations in the inertial frame are:
$$
\begin{aligned}
\dot{x}_i &= v_i \cos \phi_i, \\
\dot{y}_i &= v_i \sin \phi_i, \\
\dot{\phi}_i &= \frac{g \tan \varphi_i}{v_i},
\end{aligned}
$$
where $(x_i, y_i)$ is the position, $v_i$ the ground speed, $\phi_i$ the heading angle, $\varphi_i$ the roll angle, and $g$ the gravitational acceleration. The lateral acceleration $a_i$ is related to roll by $a_i = g \tan \varphi_i$. The control inputs are throttle (affecting $v_i$) and aileron (affecting $\dot{\phi}_i$). The following constraints apply for fixed-wing drones:
$$
v_i \in [v_{\min}, v_{\max}],\quad \varphi_i \in [\varphi_{\min}, \varphi_{\max}].
$$
Typical values for small fixed-wing drones are $v_{\min}=20\;{\rm m/s}$, $v_{\max}=75\;{\rm m/s}$, and $\varphi_{\max}=30^\circ$.
2.2 Formation Topology
We choose a diamond formation consisting of one leader (UAV1) and three followers (UAV2–UAV4). The communication topology is a strongly connected directed graph as shown in Table 1. Followers receive state information (position, speed, heading) from the leader, and also exchange positions with adjacent followers to enable distributed collision avoidance. This topology reduces dependency on the leader and mitigates error accumulation.
| Source | Destination | Information exchanged |
|---|---|---|
| UAV1 (Leader) | UAV2, UAV3, UAV4 | Position, velocity, heading |
| UAV2 | UAV3 | Position |
| UAV3 | UAV2, UAV4 | Position |
| UAV4 | UAV3 | Position |
3. Formation Control Law Design
3.1 Leader Control
The leader (UAV1) follows a predefined path using a nonlinear guidance law. A target point $P_1$ is chosen on the path at a distance $L_1$ ahead of the drone. The lateral acceleration required is:
$$
a_1 = \frac{2 v_1^2}{L_1} \sin \psi_1,
$$
where $\psi_1$ is the angle between the velocity vector $\vec{v}_1$ and the line-of-sight vector $\vec{L}_1$. This law guarantees smooth tracking of straight lines and turns.
3.2 Follower Control
Each follower (UAV $i$, $i=2,3,4$) computes its desired position relative to the leader in the leader’s body frame. Define the relative distance components:
$$
l_{X_i} = (x_i – x_1)\cos\phi_1 + (y_i – y_1)\sin\phi_1,\quad l_{Y_i} = -(x_i – x_1)\sin\phi_1 + (y_i – y_1)\cos\phi_1.
$$
The desired relative distances $(L_{X_i}, L_{Y_i})$ are fixed according to the diamond formation. The errors are $\Delta X_i = l_{X_i} – L_{X_i}$, $\Delta Y_i = l_{Y_i} – L_{Y_i}$. The follower’s speed is controlled by:
$$
v_i = v_1 – k_1 \Delta X_i,
$$
with gain $k_1>0$. The lateral acceleration is:
$$
a_i = \frac{2 v_i^2}{L_i} \sin(\alpha_i + \beta_i),
$$
where $\alpha_i = \phi_i – \phi_1$, $\beta_i = \arcsin(\Delta Y_i / L_i)$, and $L_i$ is the distance from the follower to its virtual target point. This law drives the follower toward the desired relative position while compensating for heading errors.
4. Collision Avoidance Strategy
To prevent collisions among fixed-wing drones, each follower monitors the positions of its neighbors (obtained via the distributed communication graph) and activates a temporary avoidance mode when its desired flight path enters a circular threat region around another drone. The threat region radius is set to $r_j$ (e.g., 30 m). The avoidance mechanism is triggered when the line from the follower to its virtual target intersects the threat circle.
In avoidance mode, the follower maintains a constant speed $v_0$ and uses a modified guidance law to steer around the threat. The target point is recomputed as the intersection between the tangent of the threat circle and the flight path. The lateral acceleration becomes:
$$
a_i = \frac{2 v_0^2}{L_0} \sin \psi_i,
$$
where $L_0$ is a fixed look-ahead distance. The stability of this avoidance law can be proven by analyzing the dynamics of the distance $d$ from the follower to the threat circle; it follows a second-order system:
$$
\ddot{d} + 2\xi\omega_n \dot{d} + \omega_n^2 d \approx 0,
$$
with $\xi=0.707$ and $\omega_n = \sqrt{2}v_0 c / L_0$, ensuring $d\rightarrow 0$ as $t\rightarrow\infty$ (i.e., the follower stays away from the threat). After the threat is cleared, the follower switches back to normal formation control.
5. Simulation Results
We conducted simulations to verify the formation control and collision avoidance algorithms for fixed-wing drones. The initial conditions are summarized in Table 2. The leader follows a square trajectory with a side length of 1000 m starting from (0,80) with speed 45 m/s.
| UAV | Initial position (x, y) (m) | Initial speed (m/s) | Initial heading (°) |
|---|---|---|---|
| UAV1 (Leader) | (0, 80) | 45 | 0 |
| UAV2 | (30, 130) | 45 | 0 |
| UAV3 | (-20, 130) | 45 | 0 |
| UAV4 | (10, 180) | 45 | 0 |
The results show that the fixed-wing drones maintain the diamond formation well during straight-line flight. During coordinated turns, some formation errors appear but are quickly eliminated after the turn. The root-mean-square errors in relative distances are less than 5 m in steady state.
For the collision avoidance scenario, we set the initial positions and headings as shown in Table 3. The threat radius $r_j=30$ m for all followers. The leader flies at 40 m/s.
| UAV | Initial position (x, y) (m) | Initial heading (°) |
|---|---|---|
| UAV1 (Leader) | (220, 260) | 45 |
| UAV2 | (180, 60) | 30 |
| UAV3 | (80, 20) | 60 |
| UAV4 | (30, 20) | 60 |
Without collision avoidance, UAV3 would collide with UAV2. With the proposed strategy, UAV3 successfully avoids the threat region and resumes formation flight. The minimum distance between UAV2 and UAV3 during the avoidance maneuver is 32 m, which is greater than the threat radius. The roll angles remain within the prescribed limits (max 30°). These simulations demonstrate the effectiveness of the collision avoidance algorithm for fixed-wing drones.
6. Conclusion
In this paper, we presented a cooperative formation control and collision avoidance framework specifically designed for fixed-wing drones. The leader-follower architecture combined with a distributed communication topology enables robust formation keeping and reduces communication load. The proposed control laws for the followers incorporate speed and roll angle adjustments to track the leader accurately. The collision avoidance strategy uses a threat-circle model and a temporary guidance law to prevent inter-drone collisions. Simulation results confirm that the algorithms work well under realistic constraints of fixed-wing drones, with smooth transitions between normal and avoidance modes. Future work will include event-triggered communication to further reduce bandwidth and experimental validation on actual fixed-wing drone platforms.