Anti-collision system and method based on unmanned aerial vehicle communication

Abstract

The invention belongs to the technical field of unmanned aerial vehicles, and discloses an anti-collision system and method based on unmanned aerial vehicle communication. A multi-unmanned aerial vehicle anti-collision system based on real-time communication cooperation is constructed, the physical limitation of single-point sensing is broken through, and the collision risk in a multi-unmanned aerial vehicle flight task is avoided; the system broadcasts the flight state information of the unmanned aerial vehicles in real time to realize global situation awareness beyond the visual range, and enhances the environmental awareness robustness in a complex environment in combination with a multi-source sensor data fusion technology; meanwhile, a multi-target optimization strategy and a dynamic priority strategy are adopted to realize millisecond collaborative decision-making in the cluster, an obstacle avoidance path is adjusted and optimized through multi-dimensional tracks such as speed, course and height, and safety and flight efficiency are both considered; through deep fusion of communication and intelligent decision, a systematic solution is provided for safe operation of the unmanned aerial vehicle under high-density, beyond-visual-range and complex meteorological conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicles (UAVs), and in particular relates to an anti-collision system and method based on UAV communication. Background Art

[0002] With the advancement of drone technology, drone applications are expanding beyond the military and into the civilian market, encompassing diverse areas such as logistics and distribution, power inspections, agricultural mapping, and emergency rescue. However, the risk of collisions between drones is becoming a significant concern when multiple drones are operating in coordinated flight. Currently, drone collision avoidance technology primarily relies on single-point sensing devices such as visual sensors or radar for localized perception, enabling autonomous obstacle avoidance during flight. However, the limited perception range and blind spots of individual drones in multi-drone flight create potential collision risks beyond visual range, making timely detection and response difficult.

[0003] For example, Patent No. CN118672304A discloses a collision avoidance method and system for fixed-wing swarm drones based on multi-element game theory. The patent first performs role selection and positioning and orientation analysis. Next, the leader and wingman aircraft perform collision avoidance item calculation sub-steps. The swarm control mode is then adapted, with different collision avoidance strategies implemented for different roles. Each aircraft calculates a global situation for each object it needs to avoid, generating multiple collision avoidance elements. Finally, through an update mechanism, all collision avoidance elements make game decisions to arrive at the final collision avoidance strategy. This addresses issues such as incomplete situational awareness, short collision avoidance time, and route deviations in traditional methods, thus avoiding additional risks. However, this solution relies on multi-element game theory and requires real-time calculation of the global situation and complex collision risks. This places extremely high demands on the communication bandwidth and computing power of the drone swarm, potentially leading to decision delays. This makes it difficult to ensure timely collision avoidance in large-scale swarms or high-speed dynamic scenarios. Furthermore, the fixed roles can lead to insufficient strategy flexibility, making it difficult to quickly adapt to sudden environmental changes or individual failures.

[0004] For example, Patent No. CN118798608A discloses a UAV collision avoidance method and system based on a particle swarm genetic algorithm. This patent first receives UAV dispatch instructions to obtain the UAV's electronic tag information. It then groups the dispatched UAV's electronic tags and randomly generates identification positions for the electronic tags. The electronic tags are then used as particles and the identification position of each electronic tag is globally optimized to obtain its optimal identification position using an adaptive chaotic particle swarm collision avoidance genetic algorithm. The electronic tags are then controlled to send signature information to corresponding readers based on a composite guidance vector. Based on the signature information, the electronic tags are identified and classified, and UAV registration and management are completed. Finally, the electronic tag identification positions are rationally deployed to reduce the risk of collisions between multiple UAVs and improve their safety and efficiency. However, this solution, based on the particle swarm genetic algorithm, requires multiple iterations and is prone to response lag in real-time UAV obstacle avoidance scenarios. Furthermore, the electronic tag deployment depends on the initial random positions and algorithm convergence, which can lead to local optimal solutions and fail to cover all potential collision risks in complex environments. Furthermore, the system relies on stable communication between the reader and the electronic tags, which poses a risk of recognition failure in environments with electromagnetic interference or obstruction.

[0005] Existing drone collision avoidance systems often rely on single-point sensing devices like visual sensors, radar, and lidar, as well as positioning systems like GPS and Beidou satellites. While these methods offer some obstacle avoidance capabilities, they are prone to misjudgment in complex environments and the inability to predict flight paths in advance, leading to collisions. Existing technologies also suffer from numerous adaptability issues, such as the attenuation of satellite positioning signals in obstructed environments, the impact of multipath effects on radar accuracy, and the reduced efficiency of distributed obstacle avoidance strategies as the size of the swarm increases. Therefore, a new collision avoidance system and device is urgently needed that can effectively identify and avoid potential collisions in advance. Summary of the Invention

[0006] The purpose of the present invention is to address the deficiencies in the prior art and to provide a collision avoidance system and method based on UAV communication; it aims to break through the physical limitations of single-point perception and avoid collision risks in multi-machine flight missions by constructing a multi-UAV collision avoidance system based on real-time communication collaboration. The system broadcasts its own flight status information between UAVs in real time to achieve global situational awareness beyond visual range, and combines multi-source sensor data fusion technology to enhance the robustness of environmental perception in complex environments. At the same time, multi-objective optimization strategies and dynamic priority strategies are adopted to achieve millisecond-level collaborative decision-making within the cluster, and optimize obstacle avoidance paths through multi-dimensional trajectory adjustments such as speed, heading, and altitude, taking into account both safety and flight efficiency. Through the deep integration of communication and intelligent decision-making, this solution provides a systematic solution for the safe operation of UAVs in high-density, beyond-visual-range and complex weather conditions.

[0007] In order to achieve the above-mentioned purpose, the present invention is implemented through the following technical solutions: In the first aspect, the present invention provides an anti-collision system based on UAV communication, which is used for multi-UAV anti-collision.

[0008] Each UAV is equipped with a multimodal perception fusion module, a communication module, a flight status receiving module, a data processing module, a collision risk assessment module, and an adjustment control module;

[0009] The multimodal perception fusion module integrates radar, visual sensors, satellite positioning, and IMU to collect multimodal flight status information of the UAV;

[0010] The communication module is used to regularly broadcast the drone's own flight status information and receive flight status information from other drones, and adaptively adjust the drone's own broadcast frequency, while monitoring channel quality and switching to the optimal communication channel in real time;

[0011] The flight status receiving module is used to parse the flight status information output by the communication module to ensure the reliability of the data;

[0012] The data processing module is used to fuse the flight status information related to the drone's position and trajectory after being processed by the flight status receiving module, eliminate the error of a single sensor, align the fused drone's own flight status information with the received flight status information in time and space, and construct a three-dimensional grid situation map containing the drone's own flight status information, the positions of surrounding drones, and the positions of obstacles, and mark the idle, occupied, and dynamic risk states;

[0013] The collision risk assessment module extracts the position and speed information of the UAV and the obstacle based on the flight status information after time-space alignment, thereby obtaining the relative speed component v between the UAV and the obstacle. rel , if v rel ≤0 means no collision risk, otherwise v rel >0 indicates there is a collision risk, and the collision risk assessment result is output;

[0014] The adjustment control module adopts a multi-objective optimization strategy based on the output collision risk assessment results, combined with a three-dimensional grid situation map, and dynamically adjusts the UAV flight trajectory according to the vertical height, speed, and heading priority.

[0015] Furthermore, the communication module includes:

[0016] Wireless communication network module: used to realize wireless signal transmission between drones, broadcast the drone's own flight status information, and receive the flight status information of other drones;

[0017] Information encryption and verification module: used to encrypt the broadcast flight status information and verify the received flight status information;

[0018] Channel Adaptive Module: It is used to monitor channel quality in real time, judge channel quality using signal strength detection and bit error rate analysis technology, and automatically switch to other preset channels when channel quality deteriorates;

[0019] Environmental monitoring module: used to monitor the number of surrounding drones, signal interference intensity and the complexity of the flight area, and coordinate the relationship between the three to adaptively adjust the drone's own broadcast frequency.

[0020] Furthermore, the channel adaptation module specifically adopts the logarithmic distance path loss model and BPSK modulation bit error rate to realize signal strength detection and bit error rate analysis.

[0021] The specific formula of the logarithmic distance path loss model is as follows:

[0022]

[0023] Among them, L p (d) is the path loss at the current distance d, L p (d0) is the reference path loss at the reference distance d0, n is the path loss exponent, X σ As shadows fade,

[0024] The BPSK modulation bit error rate formula is as follows:

[0025]

[0026] Where erfc(x) is the complementary error function, is the ratio of bit energy to noise spectral density;

[0027] When the signal strength is lower than the signal strength threshold or the bit error rate is higher than the bit error rate threshold, it indicates that the channel quality has deteriorated and the system automatically switches to other preset channels.

[0028] Furthermore, the details of the environmental monitoring module are as follows:

[0029]

[0030] Among them, f new 、f base and f max are the adjusted frequency, basic frequency and upper frequency limit respectively; α is the adjustment sensitivity coefficient; ω i is the weight of each parameter, i = 1, 2, 3, ω1+ω2+ω3=1; is the ratio of the number of surrounding drones to the preset maximum number; I represents the signal interference intensity, which is normalized; C represents the regional complexity.

[0031] Furthermore, it is characterized in that the flight status receiving module is used to parse the flight status information output by the communication module, specifically:

[0032] First, the flight status information output by the communication module is decoded, the flight status information is extracted, and converted into a processable data format; filtering technology is used to remove noise information, and then error correction coding technology is used to correct errors. Finally, the Kalman filter algorithm is used to process the error-corrected signal. Through the prediction and correction process, the error-corrected signal is estimated and compensated to extract accurate flight status information.

[0033] Furthermore, it is characterized in that the data processing module is specifically:

[0034] Using Kalman filter algorithm for data fusion:

[0035]

[0036] z k =Hx k +v k

[0037] Among them, x k is the state vector of the drone at the current moment, is the state transfer matrix, B is the UAV control input matrix, u k represents the drone control input vector, z k Represents the flight status information after multimodal perception fusion, H is the observation matrix, w k and v k is the noise term;

[0038] In order to eliminate the temporal and spatial errors of different data, the fused data of the UAV's own position, speed, heading, flight altitude and external obstacle information monitored by the sensor are aligned in time and space with the fused data of the position, speed, heading, flight altitude and external obstacle information received from other UAVs. The unified coordinate system is optimized through nonlinear graphs, the octree rasterization technology is used to divide the three-dimensional space into cubic grids, and pre-loaded electronic fences are used to mark the idle, occupied and dynamic risk states.

[0039] Furthermore, the collision risk assessment module calculates the remaining time to collision and divides the risk level into different levels based on the task priority. The smaller the TTC, the higher the collision risk:

[0040] The position and speed information of the UAV and the obstacle are extracted from the flight status information after time and space alignment, so as to obtain the relative distance d and relative speed component v between the UAV and the obstacle. rel , the collision time TTC is predicted according to the following formula:

[0041]

[0042] Furthermore, the dynamic adjustment of the UAV flight trajectory in the adjustment control module adopts a multi-objective optimization strategy:

[0043]

[0044] min(α·TTC+β·ΔE)

[0045] Where k1, k2 and k3 are the energy consumption coefficients of the UAV, k1+k2+k3=1; TTC is the remaining time to collision; ΔE is the energy consumption increment; α and β are weight coefficients, α+β=1; θ represents the heading; Δd represents the vertical height, and t represents the execution time;

[0046] Taking into account the two objectives of collision risk and energy consumption, the weight coefficient is adjusted and the combination that minimizes the objective function value is selected as the optimal solution, that is, the vertical height, speed, and heading adjustment decision variables are obtained.

[0047] Furthermore, the PID algorithm based on power control is used to adjust the flight speed of the drone:

[0048]

[0049] Where u(t) is the controller output at time t, i.e., the control variable on the UAV motor; e9t) is the error at time t, which is equal to the difference between the expected flight speed and the actual measured flight speed e9t) = r(t) - y(t), K p , K i , K d are the proportional, integral and differential coefficients respectively.

[0050] In a second aspect, the present invention provides a collision avoidance method based on UAV communication, which is used to implement the system, comprising the following steps:

[0051] Step S1, system initialization;

[0052] Step S2, broadcasting its own flight status information: The UAV regularly broadcasts its own flight status information through the communication module, and adaptively adjusts its own broadcast frequency based on the number of surrounding UAVs, signal interference intensity, and the complexity of the flight area. At the same time, it monitors the channel quality and switches to the optimal communication channel in real time;

[0053] Step S3, receiving and processing flight status information from other drones: parsing the flight status information broadcast by other drones, using filtering technology to remove noise signals, using error correction coding technology to correct errors, and using the Kalman filter algorithm to process the error-corrected signals. Through a prediction and correction process, the noise in the error-corrected signals is estimated and compensated, and accurate flight status information is extracted from the noisy signals.

[0054] Step S4, sensor data fusion: The data processed by the flight status receiving module is fused, and the fused flight status information of the UAV itself is aligned with the received flight status information in time and space to construct a three-dimensional grid situation map containing the UAV's own flight status information, the positions of surrounding UAVs, and the positions of obstacles, and the "idle", "occupied", and "dynamic risk" states are marked;

[0055] Step S5, collision risk assessment: extract the position and velocity information of the UAV and the obstacle based on the time-space aligned data, and obtain the relative distance d and relative velocity component v between the UAV and the obstacle. rel , according to the relative velocity component v rel Determine whether there is a collision risk, calculate the predicted collision time, and divide the risk level based on the task priority;

[0056] Step S6, dynamic trajectory adjustment: a multi-objective optimization strategy is used to balance collision risk and energy consumption, and the flight trajectory is adjusted according to the vertical height, speed, and heading priority in combination with the three-dimensional grid situation map;

[0057] Step S7, real-time feedback correction: After the dynamic trajectory is adjusted, the drone repeats steps S2-S5 to determine the collision risk in real time. If there is no risk, it continues to repeat the above steps. If there is a risk, it executes step S6 to adjust the obstacle avoidance action.

[0058] The present invention has the following beneficial effects: The UAV communication-based collision avoidance system and method provided by the present invention enables global situational awareness beyond visual range (BLOS) through real-time communication between UAVs. Combining multi-source sensor data fusion with intelligent decision-making algorithms, this overcomes the limitations of traditional single-point sensing technology, effectively improving the safety of multi-aircraft flight. In particular, it enables proactive obstacle avoidance beyond visual range (BLOS), significantly reducing the risk of flight collisions. By integrating with existing sensors, the system can provide more intelligent and accurate obstacle avoidance decisions, providing strong support for the use of UAVs in complex environments.

[0059] Adaptive broadcast strategy: The drone communication module adaptively adjusts the broadcast channel and broadcast mode in real time based on the current broadcast quality and signal strength to ensure that communication between multiple drones remains smooth and accurate at all times.

[0060] Beyond visual range obstacle avoidance capability: By regularly broadcasting flight status information, the drone does not need to rely on visual or radar detection. When it is beyond its detection range, it can obtain information about surrounding drones in advance, predict potential collision risks, and adjust the flight trajectory in advance to avoid collisions.

[0061] Multi-UAV Collaboration: The system coordinates the flight paths of multiple drones, preventing collisions between them. It also monitors and adjusts the trajectories of multiple drones in real time, avoiding collisions and improving flight safety and airspace efficiency. Multiple drones broadcast and share information, including their locations and obstacles, in real time, allowing for pre-planned flight paths and collision avoidance.

[0062] Compatible with existing technologies: Seamlessly integrates with traditional sensors, fully utilizing existing equipment capabilities to enhance system perception. This eliminates the need for extensive redevelopment or replacement of equipment, reducing costs and improving system practicality and scalability.

[0063] Intelligence and real-time performance: The system automatically determines whether to adjust the flight trajectory based on the flight plan and real-time information. Adjustments are made in real time during flight, providing rapid response and adaptability to complex and changing flight environments, ensuring flight safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flow chart of the anti-collision system based on UAV communication of the present invention;

[0065] Figure 2 This is an example of a cube grid;

[0066] Figure 3 This is the Kalman filter data fusion effect;

[0067] Figure 4 Comparison of bit error rates of three modulation methods;

[0068] Figure 5 This is a simulation diagram of multi-UAV collision avoidance;

[0069] Figure 6 This is a bird's-eye view of a multi-UAV collision avoidance simulation diagram. DETAILED DESCRIPTION

[0070] In order to make the present invention easier to understand, the present invention is further described below with reference to specific embodiments and drawings, which do not limit the present invention in any way. These embodiments and drawings are only used to illustrate the present invention and are not used to limit the scope of the present invention. Without departing from the technical solution of the present invention, any changes or modifications made to the present invention that can be easily implemented by ordinary technicians in this field will fall within the scope of the claims of the present invention.

[0071] like Figure 1As shown, a collision avoidance method based on UAV communication includes the following steps:

[0072] S1: Start (System Initialization): Before startup, the system verifies the performance of communications, sensors, and powertrain systems to ensure proper operation. It also imports no-fly zones and aircraft dynamics parameters, initializes the Kalman filter and nearby drone detection mechanism, and completes system preparation. Satellite positioning provides precise drone location data; the IMU monitors drone attitude changes; the radar measures the distance, relative speed, and direction of movement between the drone and obstacles; and the visual sensor identifies obstacle type, shape, and texture characteristics.

[0073] S2: Broadcast Flight Status: The drone regularly broadcasts its location, speed, heading, altitude, and other encrypted and verified flight status information through the communication module to ensure safe and reliable flight status. The broadcast frequency is automatically adjusted based on the number of surrounding drones, signal interference intensity, and the complexity of the flight area (increasing to 5Hz in densely populated areas and reducing to 1Hz in open areas), and the optimal communication channel is switched in real time to ensure signal quality.

[0074] Specifically: the communication module includes a wireless communication network module, an information encryption and verification module, a channel adaptation module and an environmental monitoring module; the UAV uses a wireless communication network (such as Wi-Fi, LTE, 5G, etc.) to regularly broadcast the UAV's own position, speed, heading, flight altitude and other flight status information. At the same time, when broadcasting flight status information, the information and encryption and verification encrypt the broadcast flight status information to ensure the security of information transmission and prevent the information from being stolen or tampered with; in addition, the received flight status information is verified to ensure the integrity and accuracy of the data. In addition, to ensure stable information transmission, channel adaptation technology is used: real-time monitoring of channel quality, the logarithmic distance path loss model and the BPSK modulation bit error rate formula are used to form a signal strength detection and bit error rate analysis algorithm. The specific formula of the logarithmic distance path loss model is as follows:

[0075]

[0076] Among them, L p (d) is the path loss at the current distance d, L p (d0) is the reference path loss at the reference distance d0, n is the path loss index (set according to the actual situation, such as 3.5 in urban environment and 2.5 in suburban environment), X σ Shadow fading reflects random interference such as building occlusion.

[0077] The BPSK modulation bit error rate formula is as follows:

[0078]

[0079] Where erfc(x) is the complementary error function, is the ratio of bit energy to noise spectral density.

[0080] The above method is used to monitor the channel signal strength and bit error rate in real time. Once the signal strength is lower than the threshold or the bit error rate is higher than the threshold, it indicates that the channel quality has deteriorated, and the system will automatically switch to other preset channels, including dedicated channels optimized for UAV collaborative communications, public cellular network channels with wide coverage and strong anti-interference capabilities, and dynamically extended idle channels obtained through real-time scanning, to ensure stable transmission of flight status information. These channels are prioritized according to the principle of "high reliability and low latency first": for example, the dedicated channel with strong signal, low bit error rate and low interference has the highest priority, followed by the dynamically extended low-interference and high-availability channel. The public cellular network channel as a backup channel has a lower priority than the dedicated channel, but higher than the backup dedicated channel with poor signal quality. Figure 4 As shown, in the anti-collision method based on drone communication of the present invention, the bit error rates of the three debugging methods BPSK, QPSK and QAM are compared. BPSK has the strongest anti-interference ability, ensuring that the communication between multiple drones remains smooth and accurate at all times.

[0081] The environmental monitoring algorithm automatically adjusts the broadcast frequency by monitoring factors such as the number of surrounding drones, signal interference intensity, and the complexity of the flight area. In complex environments or high-density flight areas, such as urban areas with dense high-rise buildings, where there are many drones and high signal interference, the module automatically increases the broadcast frequency to ensure timely and stable information transmission. In open areas with less interference, such as remote mountainous areas, the frequency is appropriately lowered to reduce energy consumption. The frequency adjustment is based on the monitored environmental parameters and follows a specific formula to ensure the information refresh rate in high-density scenarios:

[0082]

[0083] Among them, f new 、f base and f max are the adjusted frequency, base frequency, and upper frequency limit respectively; α is the adjustment sensitivity coefficient, which is used to control the rate at which the frequency changes with the environment; ω i is the weight of each parameter, i = 1, 2, 3, ω1+ω2+ω3=1; It is the ratio of the number of surrounding drones to the preset maximum number, reflecting the drone density ratio; I represents the signal interference intensity and is normalized; C represents the regional complexity, which is divided according to the preset level, for example, 0 = open suburbs, 1 = low-altitude urban areas, and 2 = dense buildings.

[0084] S3: Analyze other drone information: Analyze the information broadcast by surrounding drones, remove noise through anti-interference technology (filtering, error correction coding), and use the Kalman filter algorithm to compensate for signal errors to ensure data accuracy and stability. It also caches received data and retransmits or interpolates lost or erroneous data to ensure information continuity.

[0085] Specifically, advanced antenna design and signal amplification circuitry are employed to enhance the ability to capture weak signals. Furthermore, a protocol parsing algorithm (e.g., converting binary data streams or byte sequences into structured information that conforms to the protocol specifications) is used to decode the flight status information output by the communication module. Based on the corresponding UAV communication protocols for wireless communication networks (such as the dedicated UAV communication protocol Mavlink, wireless Wi-Fi (802.11), and 5G cellular networks), flight status information such as position, speed, and heading is identified and converted into a processable data format, ensuring the accuracy of the received information. This provides a reliable basis for subsequent data processing. Filtering techniques are then used to remove noise signals, and error correction is performed using error correction coding techniques to ensure stable information reception in complex electromagnetic environments and reduce data loss and errors. Finally, a Kalman filter algorithm is used to process the corrected signal. Through a prediction and correction process, the noise in the corrected signal is estimated and compensated, extracting accurate flight status information from the noisy signal and ensuring data reliability.

[0086] S4: Sensor data fusion: Integrate radar, vision, satellite positioning, IMU and other sensors to undertake core perception tasks; the satellite positioning system can provide high-precision position and altitude, the IMU (inertial measurement unit) can monitor flight attitude and acceleration, and the radar and camera are used to measure the distance and speed of obstacles and identify the type and shape of obstacles respectively. However, different sensors have their own limitations. For example, satellite signals may be lost in areas with dense buildings, and the position of the drone cannot be accurately grasped. Radar is easily interfered with by radar signals of other drones in areas with dense drones, increasing errors in data such as speed and altitude. The performance of the camera will decline in insufficient light, and the ability to perceive surrounding obstacles will decline. Therefore, it is necessary to use data fusion technology to complement the data of these sensors to eliminate the errors of a single sensor. For example Figure 3 When fusion is shown, the Kalman filter algorithm is used for data fusion:

[0087]

[0088] z k =Hx k +v k

[0089] Among them, x kis the state vector of the drone at the current moment, is the state transfer matrix, which describes the state change of a single type of data of the drone. For example, when calculating the state change of the drone speed, It also contains the drone speed data from multiple sensor sources, B is the drone control input matrix, combined with the drone control input vector u k Predict future state, z k Represents the flight state information after multimodal perception fusion, H is the observation matrix, which contains the three coordinate information of the sensor xyz. k (including position, speed, attitude, etc.) is mapped to the observation space of each sensor, w k and v k is the noise term, w k Error in response to control input, v k Reflects the error of the sensor. The first formula predicts the current state based on the state of the drone at the previous moment. The second formula describes the relationship between the sensor's measurement value and the actual state. The theoretical state is first predicted by the first formula, and then the actual observation results are obtained by the second formula. Multi-source heterogeneous data are integrated to eliminate the error of single sensor data, thereby improving the perception accuracy in complex environments. On the basis of data fusion, in order to eliminate the errors in time and space of different data, the fused data of its own position, speed, heading, flight altitude and external obstacle information monitored by its own sensors are aligned with the fused data of the position, speed, heading, flight altitude and external obstacle information received from other drones, as well as the drone density, signal interference intensity and regional complexity received by the communication module. That is, the time difference of different drones for the same information is eliminated, ensuring that the information during drone communication is synchronized on the time scale, and the unified coordinate system is optimized through nonlinear graphs to eliminate the pose / target coordinate error. The octree rasterization technology is used to divide the three-dimensional space into Figure 2 The cube grid shown uses pre-loaded geo-fences to mark "free", "occupied" and "dynamic risk" states.

[0090] S5: Collision risk assessment: Extract the position and velocity information of the UAV and the obstacle based on the time-space aligned data, and then obtain the relative distance d and relative velocity component v between the UAV and the obstacle. rel (i.e. the shortest distance between the two vehicles and the collision), the time to collision (TTC) is predicted according to the following formula:

[0091]

[0092] If v rel ≤0 means no collision risk, otherwise v rel When >0, it indicates a collision risk. Calculate the ratio v of the relative distance d to the relative velocity component.rel , that is, the remaining time for collision TTC. The risk level is divided in combination with the task priority, and the collision risk assessment result is output. For example, based on the historical trajectory, the future flight path is predicted, the potential collision path is identified and a "conflict time window" is generated, and high-risk conflicts (such as TTC≤5s) are handled first. In addition, even if there is no collision risk through assessment, the drone will combine the "idle", "occupied" and "dynamic risk" status marked in real time by the pre-loaded electronic fence in step three to judge in advance whether there is a potential risk in the area ahead and readjust the route. For example, the "idle" area has a lower potential collision risk and can fly for a long time. The "occupied" area has a greater potential collision risk and needs to be evacuated as soon as possible. The "dynamic risk" area is between the two. It is necessary to judge whether to evacuate based on actual needs, and choose to detour if it is not necessary.

[0093] S6: Dynamic trajectory adjustment: When a collision risk is confirmed, based on the risk assessment results and a three-dimensional grid situation map, the flight trajectory is adjusted according to the priority of "vertical height → speed → heading" (limiting the height difference between drones to ≥30m, speed adjustment to ±10m / s, and heading angle to ±30°). This multi-objective optimization decision-making balances safety and energy consumption. In high-risk situations, emergency braking or hovering is initiated, and in the event of sensor failure, a return to home is triggered to avoid danger, ensuring safety in extreme situations.

[0094] Within the physically feasible range, the system traverses different combinations of the three decision variables of trajectory heading θ, vertical height Δd, and execution time t, and the multi-objective optimization function determines the UAV trajectory adjustment strategy:

[0095]

[0096] min(α·TTC+β·ΔE)

[0097] Where k1, k2, and k3 are the UAV's energy consumption coefficients, k1 + k2 + k3 = 1; TTC is the time to collision, ΔE is the incremental energy consumption, and α and β are weighting coefficients, α + β = 1. This algorithm comprehensively considers both collision risk and energy consumption. By adjusting the weighting coefficients, it balances the importance of both based on actual needs and selects the combination that minimizes the objective function value as the optimal solution. In practical applications, if the current flight mission has a high time requirement, the weight of α can be appropriately increased to prioritize shortening the time to collision. If energy efficiency is a priority, the weight of β can be increased to reduce energy consumption. This solution is ultimately converted into specific trajectory parameters to guide the UAV's flight adjustments. The multi-objective optimization function, combined with a three-dimensional grid situation map, can determine the physically feasible range of the three UAV variables. For example, regarding the course, whether surrounding obstacles allow the UAV to make sufficient steering angles without collision.

[0098] Adjust the control module and use the PID algorithm based on power control to change the flight speed:

[0099]

[0100] Where u(t) is the controller output at time t, i.e., the control variable on the motor of the drone. e(t) is the error at time t, which is equal to the difference between the expected flight speed and the actual measured flight speed. e(t) = r(t) - y(t), K p , K i , K d These are the proportional, integral, and differential coefficients, respectively. This algorithm calculates u(t) to adjust the speed of the drone's motor, allowing the actual speed of the drone to approach the desired speed, achieving fast and smooth acceleration and deceleration. When changing course, in addition to using the PID algorithm, an attitude control-based steering algorithm can also be used, combined with attitude data measured by the gyroscope to accurately control the steering angle.

[0101] When changing the height, the height control algorithm is used to adjust the lifting power according to the difference between the set height and the actual height to ensure smooth ascent or descent.

[0102] S7: Real-time feedback correction: After the dynamic trajectory is adjusted, the drone repeats steps S2-S5, and judges the collision risk in real time based on the sensor feedback. If there is no risk, it continues to repeat the above steps. If there is a risk, it executes step S6 to adjust the obstacle avoidance action. First, the trajectory adjustment strategy is determined through the multi-objective optimization function, and then the PID algorithm is used to control the flight speed. The actual speed is close to the expected speed through the motor control amount. The attitude control algorithm is combined with the altitude feedback control algorithm for precise steering and the altitude feedback control algorithm for smooth ascent and descent. In extreme cases, emergency braking or hovering will be triggered to adjust the obstacle avoidance action, smooth the trajectory and compensate for errors to ensure a smooth and efficient obstacle avoidance process.

[0103] like Figure 5 and Figure 6 As shown in the figure, the multi-objective optimization function combined with the three-dimensional grid situation map can realize the real-time broadcasting of multiple UAVs to share information such as location and obstacles, plan flight routes in advance, and avoid collision risks.

[0104] An anti-collision system based on UAV communication, each UAV is equipped with a multimodal perception fusion module, a communication module, a flight status receiving module, a data processing module, a collision risk assessment module and an adjustment control module;

[0105] The multimodal perception fusion module integrates radar, visual sensors, satellite positioning, and IMU. Satellite positioning can provide accurate location data for the drone; the IMU can monitor the drone's attitude changes; the radar can measure the distance, relative speed, and direction of movement between the drone and obstacles; and the visual sensor can identify the type, shape, and texture characteristics of obstacles.

[0106] The communication module is used to regularly broadcast the drone's own flight status information and receive flight status information from other drones, and adaptively adjust its own drone's broadcast frequency. At the same time, it monitors the channel quality and switches to the optimal communication channel in real time.

[0107] The flight status receiving module is used to parse the flight status information output by the communication module to ensure the reliability of the data;

[0108] The data processing module is used to fuse the data related to the drone's position and trajectory processed by the flight status receiving module, eliminate the error of a single sensor, align the fused drone's own flight status information with the received flight status information in time and space, and construct a three-dimensional grid situation map that includes the drone's own flight status information, the positions of surrounding drones, and the positions of obstacles, and mark the idle, occupied, and dynamic risk states;

[0109] This module is the "intelligent brain" of the entire system. After receiving flight status information, it conducts an in-depth analysis based on its own flight plan (including starting point, waypoints, destination, and flight speed planning). Using a machine learning-based collision risk prediction algorithm, it comprehensively considers factors such as the drone's flight speed, direction, and position change trends to accurately calculate potential collision risks and assess the likelihood and timing of collisions. This module deeply integrates the flight status information received from other drones with its own flight plan (including starting point, waypoints, destination, and flight speed planning). Using a data association algorithm, it matches and integrates data from different sources to form a unified dataset for subsequent analysis.

[0110] Before a collision occurs, this module implements a collision risk prediction algorithm by incorporating machine learning algorithms, such as neural networks. This algorithm is trained using extensive historical flight data to learn the relationship between factors such as the drone's flight speed, direction, and position trends, and collision risk. During actual operation, the module inputs flight data for the current drone and surrounding drones to predict potential collision risks and assess the likelihood and timing of a collision. For example, based on multiple sets of historical data on the relative positions and speeds of drones, as well as the eventual collision outcome, the neural network model is trained to accurately determine collision risk in the current flight state.

[0111] The collision risk assessment module extracts the position and speed information of the UAV and the obstacle based on the flight status information after time and space alignment, thereby obtaining the relative speed component v between the UAV and the obstacle. rel , if v rel ≤0 means no collision risk, otherwise v rel >0 indicates there is a collision risk, and the collision risk assessment result is output;

[0112] The adjustment control module dynamically adjusts the UAV flight trajectory according to the vertical height, speed, and heading priority based on the output collision risk assessment results and the three-dimensional grid situation map.

[0113] Once the data processing module determines a collision risk, the adjustment control module immediately activates. Based on the risk assessment results, the drone's flight trajectory is automatically and accurately adjusted. Within the physically feasible range, the system traverses different combinations of the three decision variables: trajectory adjustment direction θ, vertical height Δd, and execution time t. It then determines the drone's trajectory adjustment strategy using a multi-objective optimization function:

[0114] In extremely dangerous situations, the module can also perform emergency braking and hovering. During emergency braking, power output is quickly cut off and the brakes are activated, bringing the drone to a complete stop. The hovering function relies on advanced attitude control algorithms and power balancing technology to maintain a stable hover and avoid collisions. This algorithm adjusts the output power of each motor in real time to offset external interference and maintain the drone's stable attitude.

[0115] The above shows and describes the basic principles, main features, and advantages of the present invention. However, the above is only a specific embodiment of the present invention, and the technical features of the present invention are not limited thereto. Any other implementation methods derived by any person skilled in the art without departing from the technical solution of the present invention should be included in the patent scope of the present invention.

Claims

1. A collision avoidance system based on drone communication, used for multi-drone collision avoidance, characterized in that: Each UAV is equipped with a multimodal perception fusion module, a communication module, a flight status receiving module, a data processing module, a collision risk assessment module, and an adjustment control module; The multimodal perception fusion module integrates radar, visual sensors, satellite positioning, and IMU to collect multimodal flight status information of the UAV; The communication module is used to regularly broadcast the drone's own flight status information and receive flight status information from other drones, and adaptively adjust the drone's own broadcast frequency, while monitoring channel quality and switching to the optimal communication channel in real time; The flight status receiving module is used to parse the flight status information output by the communication module to ensure the reliability of the data; The data processing module is used to fuse the flight status information related to the drone's position and trajectory after being processed by the flight status receiving module, eliminate the error of a single sensor, align the fused drone's own flight status information with the received flight status information in time and space, and construct a three-dimensional grid situation map containing the drone's own flight status information, the positions of surrounding drones, and the positions of obstacles, and mark the idle, occupied, and dynamic risk states; The collision risk assessment module extracts the position and speed information of the UAV and the obstacle based on the flight status information after time-space alignment, thereby obtaining the relative speed component v between the UAV and the obstacle. rel , if v rel ≤0 means no collision risk, otherwise v rel >0 indicates there is a collision risk, and the collision risk assessment result is output; The adjustment control module adopts a multi-objective optimization strategy based on the output collision risk assessment results, combined with a three-dimensional grid situation map, and dynamically adjusts the UAV flight trajectory according to the vertical height, speed, and heading priority.

2. The anti-collision system based on drone communication according to claim 1, characterized in that: The communication module includes: Wireless communication network module: used to realize wireless signal transmission between drones, broadcast the drone's own flight status information, and receive the flight status information of other drones; Information encryption and verification module: used to encrypt the broadcast flight status information and verify the received flight status information; Channel Adaptive Module: It is used to monitor channel quality in real time, judge channel quality using signal strength detection and bit error rate analysis technology, and automatically switch to other preset channels when channel quality deteriorates; Environmental monitoring module: used to monitor the number of surrounding drones, signal interference intensity and the complexity of the flight area, and coordinate the relationship between the three to adaptively adjust the drone's own broadcast frequency.

3. The anti-collision system based on UAV communication according to claim 2, characterized in that: The channel adaptation module specifically uses the logarithmic distance path loss model and BPSK modulation bit error rate to achieve signal strength detection and bit error rate analysis. The specific formula of the logarithmic distance path loss model is as follows: Among them, L p (d) is the path loss at the current distance d, L p (d0) is the reference path loss at the reference distance d0, n is the path loss exponent, X σ As shadows fade, The BPSK modulation bit error rate formula is as follows: Where erfc(x) is the complementary error function, is the ratio of bit energy to noise spectral density; When the signal strength is lower than the signal strength threshold or the bit error rate is higher than the bit error rate threshold, it indicates that the channel quality has deteriorated and the system automatically switches to other preset channels.

4. The anti-collision system based on UAV communication according to claim 2, characterized in that: The details of the environmental monitoring module are as follows: Among them, f new 、f base and f max are the adjusted frequency, basic frequency and upper frequency limit respectively; α is the adjustment sensitivity coefficient; ω i is the weight of each parameter, i = 1, 2, 3, ω1+ω2+ω3=1; is the ratio of the number of surrounding drones to the preset maximum number; I represents the signal interference intensity, which is normalized; C represents the regional complexity.

5. The anti-collision system based on drone communication according to claim 1, characterized in that: The flight status receiving module is used to parse the flight status information output by the communication module, specifically: First, the flight status information output by the communication module is decoded, the flight status information is extracted, and converted into a processable data format; filtering technology is used to remove noise information, and then error correction coding technology is used to correct errors. Finally, the Kalman filter algorithm is used to process the error-corrected signal. Through the prediction and correction process, the error-corrected signal is estimated and compensated to extract accurate flight status information.

6. The anti-collision system based on UAV communication according to claim 1, characterized in that: The data processing module is specifically: Using Kalman filter algorithm for data fusion: z k =Hx k +v k Among them, x k is the state vector of the drone at the current moment, is the state transfer matrix, B is the UAV control input matrix, u k represents the drone control input vector, z k Represents the flight status information after multimodal perception fusion, H is the observation matrix, w k and v k is the noise term; In order to eliminate the temporal and spatial errors of different data, the fused data of the UAV's own position, speed, heading, flight altitude and external obstacle information monitored by the sensor are aligned in time and space with the fused data of the position, speed, heading, flight altitude and external obstacle information received from other UAVs. The unified coordinate system is optimized through nonlinear graphs, the octree rasterization technology is used to divide the three-dimensional space into cubic grids, and pre-loaded electronic fences are used to mark the idle, occupied and dynamic risk states.

7. The anti-collision system based on UAV communication according to claim 1, characterized in that: The collision risk assessment module calculates the remaining time to collision and divides the risk level into two levels based on the task priority. The smaller the TTC, the higher the collision risk: The position and speed information of the UAV and the obstacle are extracted from the flight status information after time and space alignment, so as to obtain the relative distance d and relative speed component v between the UAV and the obstacle. rel , the collision time TTC is predicted according to the following formula:

8. The anti-collision system based on UAV communication according to claim 1, characterized in that: The dynamic adjustment of the UAV flight trajectory in the adjustment control module adopts a multi-objective optimization strategy: min(α·TTC+β·ΔE) Where k1, k2 and k3 are the energy consumption coefficients of the UAV, k1+k2+k3=1; TTC is the remaining time to collision; ΔE is the energy consumption increment; α and β are weight coefficients, α+β=1; θ represents the heading; Δd represents the vertical height, and t represents the execution time; Taking into account the two objectives of collision risk and energy consumption, the weight coefficient is adjusted and the combination that minimizes the objective function value is selected as the optimal solution, that is, the vertical height, speed, and heading adjustment decision variables are obtained.

9. The anti-collision system based on drone communication according to claim 8, characterized in that: Use the PID algorithm based on power control to adjust the flight speed of the drone: Where u(t) is the controller output at time t, i.e., the control variable on the UAV motor; e(t) is the error at time t, which is equal to the difference between the expected flight speed and the actual measured flight speed e(t) = r(t) - y(t), K p , K i , K d are the proportional, integral and differential coefficients respectively.

10. A collision avoidance method based on UAV communication, used to implement the system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Step S1, system initialization; Step S2, broadcasting its own flight status information: The UAV regularly broadcasts its own flight status information through the communication module, and adaptively adjusts its own broadcast frequency based on the number of surrounding UAVs, signal interference intensity, and the complexity of the flight area. At the same time, it monitors the channel quality and switches to the optimal communication channel in real time; Step S3, receiving and processing flight status information from other drones: parsing the flight status information broadcast by other drones, using filtering technology to remove noise signals, using error correction coding technology to correct errors, and using the Kalman filter algorithm to process the error-corrected signals. Through a prediction and correction process, the noise in the error-corrected signals is estimated and compensated, and accurate flight status information is extracted from the noisy signals. Step S4, sensor data fusion: The data processed by the flight status receiving module is fused, and the fused flight status information of the UAV itself is aligned with the received flight status information in time and space to construct a three-dimensional grid situation map containing the UAV's flight status information, the positions of surrounding UAVs, and the positions of obstacles, and annotating the "idle", "occupied", and "dynamic risk" states; Step S5, collision risk assessment: extract the position and velocity information of the UAV and the obstacle based on the time-space aligned data, and obtain the relative distance d and relative velocity component v between the UAV and the obstacle. rel , according to the relative velocity component v rel Determine whether there is a collision risk, calculate the predicted collision time, and divide the risk level based on the task priority; Step S6, dynamic trajectory adjustment: a multi-objective optimization strategy is used to balance collision risk and energy consumption, and the flight trajectory is adjusted according to the vertical height, speed, and heading priority in combination with the three-dimensional grid situation map; Step S7, real-time feedback correction: After the dynamic trajectory is adjusted, the drone repeats steps S2-S5 to determine the collision risk in real time. If there is no risk, it continues to repeat the above steps. If there is a risk, it executes step S6 to adjust the obstacle avoidance action.

Images