Forest canopy penetration communication relay unmanned aerial vehicle system based on RISC-V architecture

Abstract

A forest canopy penetration communication relay unmanned aerial vehicle system based on an RISC-V framework comprises an unmanned aerial vehicle platform, a ground sensor node, a satellite / base station link and a main control chip based on the RISC-V framework. By expanding an RISC-V instruction set and a hardware acceleration module, dynamic channel estimation, real-time beam forming and adaptive LDPC coding are realized, and the stability and real-time performance of a communication link are remarkably improved. The forest canopy light transmittance is modeled in real time through the LiDAR technology, a low-power-consumption mode and a high-performance mode are dynamically switched, and the energy efficiency of a communication link is optimized. Through the self-adaptive beam forming and bionic antenna design, the attenuation of the forest canopy to the signal is effectively reduced, and the signal penetration capability is improved. The method has efficient dynamic resource management capability, can stably operate in a complex and dynamically changing forest environment, and is widely applied to emergency communication scenes such as forest fire early warning, ecological environment monitoring and post-disaster recovery.

Description

Technical Field

[0001] The present invention relates to the cross - field of smart forestry and wireless communication, in particular to a forest canopy penetration communication relay UAV system based on the RISC - V architecture. Background Art

[0002] Wireless communication in forest environments has long faced three major technical problems: weak signal penetration, poor dynamic adaptability, and energy efficiency imbalance. In dense canopy areas, signals in the 2.4GHz / 5GHz frequency bands are attenuated by up to 20 - 30dB due to leaf absorption, branch scattering, and multipath effects. Although traditional UAV relay systems (LoRa - based solutions) improve the transmission distance through spread - spectrum technology, their fixed spreading factors and static coding strategies are difficult to adapt to the dynamic changes of forest channels. The measured bit error rate ≥ 10 -4 , which cannot meet the requirements of high - reliability scenarios such as forest fire warning. At the same time, the algorithms and hardware of existing solutions are separated. Dynamic channel coding (such as LDPC code - length adaptation) and beamforming weight calculation rely on software implementation. The general - purpose computing units of the ARM architecture cannot efficiently process dense operations such as matrix inversion (complexity O(n 3 ))), resulting in insufficient real - time performance; at the same time, energy efficiency management is extensive, lacking a dynamic resource allocation mechanism based on environmental perception. UAVs still operate in full - power mode in sparse canopy areas, causing energy waste.

[0003] The existing patent CN114189824A proposes a UAV inspection system based on dual LoRa nodes, which reduces interference through sub - channel design (channel 0 and channel 4). However, this solution does not optimize the signal modulation method for canopy penetration characteristics and still uses the LoRa fixed spreading mechanism, unable to cope with canopy dynamic attenuation; moreover, it relies on a Raspberry Pi to process multi - sensor data, and the single - core performance of the ARM Cortex - A72 architecture used by the Raspberry Pi's processor is limited, easily causing beamforming delay. This system also does not build a linkage mechanism between the channel model and the coding strategy, resulting in a bit error rate as high as 10 - 3 in dense forest areas. In the case of high canopy density, the data packet loss rate of this solution increases significantly.

[0004] The existing publication number US20220173810A1 uses a UAV equipped with a LoRa module as a relay node, adopts the general LoRa protocol and the ARM processor architecture, and achieves low - power communication. However, its rigid physical layer parameters (fixed spreading factor SF = 12) require increasing the transmission power to maintain the link under dense canopies, resulting in a sharp reduction in the UAV's flight time to, and the general - purpose processor is difficult to efficiently support real - time beamforming and dynamic channel estimation, with an end - to - end transmission delay exceeding 100ms. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a forest canopy penetration communication relay UAV system based on the RISC-V architecture, aiming to solve the technical problems of serious communication signal attenuation, poor real-time performance, and high energy consumption resulting in short endurance time in the prior art in a dense forest environment, so as to improve the communication stability, data processing efficiency, and operation endurance ability of the UAV in a complex forest environment.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A forest canopy penetration communication relay UAV system based on the RISC-V architecture, including a UAV platform, a ground sensor node, and a satellite / base station communication link;

[0008] The UAV platform is equipped with a RISC-V main control chip, an adaptive radio frequency communication module, and a bionic camouflage shell, and is used to realize the real-time establishment and dynamic adjustment of the forest canopy penetration communication link;

[0009] The ground sensor node is used to collect forest environment information and interact with the UAV through a narrowband communication protocol;

[0010] The satellite / base station communication link is used for data forwarding between the UAV and the upper-level network.

[0011] The RISC-V main control chip adopts a 64-bit dual-core processor, with a built-in hardware acceleration module and a custom instruction set extension; among them, the hardware acceleration module supports a custom vector and matrix processing extension instruction set,

[0012] The custom instruction set extension includes the vbf.capon instruction and the crc.ldpc instruction; the vbf.capon instruction is used to optimize the communication beam direction in real time, and the crc.ldpc instruction is used for channel adaptive modulation.

[0013] The vbf.capon instruction integrates a dedicated circuit for matrix inversion based on Cholesky decomposition and supports real-time beamforming weight calculation; the formula for real-time beamforming weight calculation is W = R -1 a(θ); where R -1 is the covariance matrix, a(θ) is the beam direction angle, θ is the direction angle of beamforming, that is, the angle pointed by the main lobe of the beam; W represents the weight vector of beamforming.

[0014] The crc.ldpc instruction dynamically adjusts the LDPC code length to match the best coding rate under the ITU-R F.1819 channel conditions in real time; the formula for dynamic adjustment of the LDPC coding rate is k ∈ [256, 1536], n ∈ [512, 2048]; where: K is the information bit length, n is the total code length, and R is the LDPC coding rate.

[0015] The hardware acceleration module includes an MMSE coprocessor and a butterfly bionic flexible antenna array; the MMSE coprocessor is a channel estimation unit based on the minimum mean square error algorithm, integrating a channel matrix tracking and Doppler frequency shift compensation circuit; the butterfly bionic flexible antenna array adjusts the phase of the array antenna through a PWM signal controlled by a RISC-V processor.

[0016] The adaptive radio frequency communication module is designed based on the AD9361 chip, supports dynamic frequency hopping in the wide frequency band of 700 MHz to 6 GHz, and has a spectrum sensing and frequency hopping mechanism.

[0017] The surface of the bionic camouflage shell outside the UAV fuselage is covered with a chlorophyll spectral reflection coating.

[0018] A dynamic resource management module is set in the UAV system, which combines with the airborne LiDAR module to perform real-time modeling on the light transmittance of the forest canopy, and intelligently adjusts the processor frequency, modulation method and transmission power according to the model results.

[0019] The dynamic resource management module is a forest canopy density recognition mechanism based on lidar point cloud, which uses the light transmittance to judge the current channel environment and dynamically allocate system resources; when the light transmittance ≥ 30%, the UAV system enables the low-power mode to ensure the basic reachability of communication; when the light transmittance < 30%, the UAV system automatically activates the MMSE coprocessor, LDPC dynamic coding module and 16QAM modulation method to maximize the performance of the communication link.

[0020] The calculation formula of the light transmittance is where N ground is the number of laser points penetrating the canopy and reaching the ground, and N total is the total number of emitted laser points.

[0021] The present invention provides a forest canopy penetration communication relay UAV system based on the RISC-V architecture, having the following technical effects:

[0022] 1). Improve signal penetration performance: By combining the dedicated instruction set of the RISC-V architecture with the hardware acceleration module, the present invention significantly reduces the attenuation effect of the forest canopy on the communication signal. In the traditional communication system, the signal penetration loss in the dense forest area can reach more than 20 dB, while the system of the present invention reduces the canopy penetration loss to ≤ 10 dB through the adaptive beamforming technology and bionic antenna design, significantly improving the reliability and stability of communication.

[0023] 2) Improve real-time performance and computing power: The present invention uses a RISC-V processor and a dedicated hardware module, which supports dynamic channel estimation and real-time beamforming, reducing the latency problem caused by the insufficient computing power of general-purpose processors. By extending the RISC-V instruction set (such as vbf.capon instruction, crc.ldpc instruction), the computing latency is reduced from 12 ms in the traditional scheme to 2.3 ms, greatly improving the real-time performance of the system and adapting to the dynamically changing environment.

[0024] 3) Significantly reduce system power consumption and extend the UAV endurance: The present invention dynamically switches between low-power mode and high-performance mode based on the real-time modeling of canopy transmittance by the LiDAR module according to environmental conditions. Under sparse canopy conditions, the system enables the low-power mode (CPU frequency 100 MHz, QPSK modulation, transmit power 10 dBm), and the energy efficiency ratio is increased by 40%. The UAV endurance time is increased from the conventional 90 minutes to 120 minutes. Even in complex environments, the UAV can maintain a long operation time, suitable for long-term and high-efficiency task requirements such as forest fire risk monitoring.

[0025] 4) Adaptive dynamic resource management to optimize the communication link: The present invention realizes the efficient operation of the system by real-time detecting the canopy transmittance and dynamically adjusting communication and computing resources. When the canopy transmittance is low (such as below 30%), the system will automatically switch to the high-performance mode, start the coprocessor and 16QAM modulation method, and increase the transmit power, so as to ensure reliable communication in complex environments. This adaptive resource management mode provides a more flexible solution in complex and rapidly changing environments, further improving the robustness and reliability of the system.

[0026] 5) Suitable for various emergency and ecological monitoring scenarios: The present invention is not only applicable to emergency communication scenarios such as real-time forest fire warning, but also can be widely used in tasks such as ecological monitoring, post-disaster recovery, and wildlife tracking. Due to its strong adaptability and high stability in dynamic environments, the UAV system of the present invention can achieve stable remote data transmission and real-time information sharing in various harsh environments, greatly improving the efficiency of emergency response and ecological protection.

[0027] 6) Efficient hardware and software co-optimization: Through the close co-design of hardware and software, the present invention combines hardware acceleration modules with customized algorithms, avoiding the bottleneck problems caused by insufficient processing power or excessive dependence on general-purpose hardware in traditional communication systems. Whether in key links such as signal processing, data compression, or beamforming, the present invention achieves a balance between performance and energy efficiency, ensuring the high efficiency and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention will be further described below in conjunction with the drawings and embodiments:

[0029] Figure 1 This is the system architecture diagram of the present invention.

[0030] Figure 2 This is the schematic diagram of the custom instruction extension of the RISC-V chip in the present invention.

[0031] Figure 3 This is the mechanical structure diagram of the bionic antenna array in the present invention.

[0032] Figure 4 This is the dynamic resource management flowchart in the present invention.

[0033] Figure 5 This is the data flow diagram in the present invention. Detailed implementation manners

[0034] Example 1

[0035] As Figure 1 shown, a forest canopy penetration communication relay UAV system based on the RISC-V architecture includes a UAV platform, a ground sensor node, a satellite / base station communication link, and a dynamic resource management module.

[0036] The UAV platform is equipped with a RISC-V main control chip, an adaptive radio frequency communication module, a bionic camouflage shell, a LiDAR module, and a bionic flexible antenna array, and is used to establish and dynamically adjust the communication link in real time.

[0037] The ground sensor node interacts with the UAV through the NB-IoT protocol to collect data such as temperature and humidity, soil moisture, and CO2 concentration.

[0038] The satellite / base station communication link uses a Ka-band module or Starlink protocol equipment to realize data forwarding between the UAV and the remote command center.

[0039] The dynamic resource management module, the dynamic resource management module is a forest canopy density recognition module based on LiDAR point cloud. This module uses the canopy transmittance to judge the current channel environment and dynamically allocate system resources. When the transmittance ≥ 30% (sparse canopy), the system enables the low-power mode: the RISC-V main frequency drops to 100 MHz, QPSK modulation is used, and the transmit power is controlled within 10 dBm to ensure the basic reachability of communication; when the transmittance < 30% (dense canopy), the system automatically activates the MMSE coprocessor, the LDPC dynamic coding module, and the 16QAM modulation method, and the transmit power is increased to 20 dBm to maximize the performance of the communication link.

[0040] When the UAV system works, it executes the following work process:

[0041] S1: The drone locates the take-off point through the GPS / Beidou dual-mode navigation module, and the onboard LiDAR module performs 3D modeling of the forest canopy at a scanning rate of 100,000 points per second, calculating the light transmittance in real time Where N ground is the number of laser points that penetrate the canopy and reach the ground, N total is the total number of laser points emitted.

[0042] S2: The RISC-V main control chip dynamically switches the working mode according to the transmittance threshold (T≥30% for sparse canopy, T<30% for dense canopy). Specifically, it includes: adjusting the processor main frequency (100MHz~1.5GHz), selecting the modulation mode (QPSK or 16QAM), controlling the transmission power (10dBm~20dBm) and activating the hardware acceleration module (such as MMSE coprocessor);

[0043] S3: The adaptive RF communication module dynamically selects the 700MHz~6GHz frequency band based on the AD9361 chip, and combines spectrum sensing technology to avoid interference frequency bands; the bionic flexible antenna array adjusts the phase through the PWM signal to achieve ±60° beam scanning, focusing signal energy to improve penetration capability;

[0044] S4: The data collected by the ground sensor nodes are transmitted to the drone via the NB-IoT protocol. The RISC-V main control chip dynamically adjusts the LDPC code length (512 to 2048 bits) through the crc.ldpc instruction. After the data is encoded, it is forwarded to the command center via the satellite / base station communication link. The end-to-end transmission delay is ≤50ms.

[0045] S5: After the mission is completed, the drone returns autonomously based on the A* path planning algorithm. Before landing, the self-test program verifies the hardware status (including antenna impedance, battery capacity and chip temperature). In case of abnormality, a warning is triggered or the landing is aborted.

[0046] Preferably, the RISC-V main control chip adopts the Pingtou Ge Yiying 1520 processor (64-bit dual-core architecture, main frequency 1.5GHz), with built-in hardware acceleration module and custom instruction set extension.

[0047] like Figure 2 As shown, the instruction set extensions include the vbf.capon instruction, the crc.ldpc instruction, and the vector processing instruction set (RV64V).

[0048] Furthermore, the vbf.capon instruction integrates a dedicated circuit for matrix inversion based on Cholesky decomposition, supporting real-time beamforming weight calculation. In traditional solutions, beamforming weight calculation takes 12ms, but after hardware acceleration through the vbf.capon instruction set, the delay is reduced to 2.3ms, and the computing efficiency is improved by 80%.

[0049] Preferably, the formula for calculating real-time beamforming weights is W = R -1 a(θ); where R -1 is the covariance matrix, a(θ) is the beam direction angle, and θ is the direction angle of beamforming.

[0050] Furthermore, the crc.ldpc instruction dynamically adjusts the LDPC code length (512 - 2048 bits) to match the optimal coding rate (1 / 2 - 3 / 4) under the ITU-R F.1819 channel conditions. Tests show that in a dense canopy environment, the bit error rate drops from 10 -4 to 10 -6 .

[0051] Preferably, the formula for dynamically adjusting the LDPC coding rate is k ∈ [256, 1536], n ∈ [512, 2048]. Where: K is the information bit length, n is the total code length, and R is the LDPC coding rate.

[0052] Furthermore, the vector processing instruction set (RV64V) includes accelerating channel matrix operations and Doppler frequency shift compensation, increasing the CSI update frequency of the MMSE co-processor from 50 Hz to 100 Hz, and significantly improving the dynamic channel estimation accuracy.

[0053] Preferably, the channel matrix operation estimation formula of the MMSE co-processor is H^ = (X H X + σ 2 I) -1 X H Y. Where H^ is the channel matrix, X is the pilot signal matrix of the known training sequence transmitted, Y is the signal matrix received containing channel distortion and noise, and σ 2 is the noise power based on the variance of Gaussian white noise. X H is the conjugate transpose matrix of X, and I is the identity matrix with the same dimension as X H X.

[0054] The hardware acceleration module includes an MMSE co-processor and a butterfly-shaped bionic flexible antenna array. The MMSE co-processor is based on the least mean square error algorithm and integrates a channel matrix tracking circuit to support dynamic channel estimation; the butterfly-shaped bionic flexible antenna array consists of 8 flexible radiation units, and the unit spacing (5 mm - 30 mm) and phase difference (0° - 180°) are adjusted by the PWM signal output by the RISC-V processor to achieve dynamic adjustment of the beam width (5° - 30°). In a dense canopy environment, the antenna switches to the narrow beam mode (5°), and the signal gain is increased by 6 dB; in the sparse area, the wide beam mode (30°) is enabled, and the coverage range is expanded to a radius of 1.5 m.

[0055] The adaptive radio frequency communication module is designed based on the AD9361 chip, supports wide-band dynamic frequency hopping in the range of 700 MHz to 6 GHz, and combines spectrum sensing technology to achieve interference avoidance. Tests show that in the 2.4 GHz frequency band, the penetration loss is reduced by 40% compared with the traditional LoRa solution, and the average signal strength is increased from -85 dBm to -65 dBm.

[0056] Preferably, the bionic camouflage shell adopts a bionic camouflage structure, and its surface is covered with a chlorophyll spectral reflection coating. The coating composition is a complex of sodium copper chlorophyllin and nano-titanium oxide, with a reflectivity ≤ 5% in the visible light band and a 30% reduction in the infrared radiation intensity, effectively avoiding detection by visible light / infrared reconnaissance equipment.

[0057] Preferably, when the adaptive radio frequency communication module is working, the spectrum sensing module adopts a joint algorithm of energy detection and cyclostationary feature detection to achieve high-sensitivity detection of -110 dBm in the 6 GHz frequency band. The frequency hopping mechanism supports dynamic interval adjustment (10 - 50 ms), combined with real-time channel quality assessment, effectively avoiding co-channel interference and burst noise. The frequency offset compensation module adopts fast locking technology based on a phase-locked loop (PLL) to achieve high-precision frequency offset correction of ±50 Hz, ensuring the stability of carrier synchronization; this module adopts a Doppler compensation algorithm, uses the FFT algorithm for RISC-V hardware acceleration, and real-time tracks the frequency offset caused by the high-speed movement of the drone (≥20 m / s).

[0058] Furthermore, the bionic flexible antenna array uses a polyimide flexible substrate, and its morphology simulates the scale structure of a butterfly wing, and is deformed by a microelectromechanical system (MEMS) control unit. In a dense canopy environment, the antenna switches to the high-frequency mode (5.8 GHz), and the beam width is compressed to 5°, focusing energy to penetrate branches and leaves; in a sparse canopy environment, it switches to the low-frequency mode (900 MHz), and the beam expands to 30°, achieving wide-area coverage. Actual measurements show that this design increases the signal penetration rate by 25% and reduces the bit error rate by 50%.

[0059] As Figure 4 shown, the system uses the LiDAR module to model the canopy light transmittance in real time and dynamically switches the working mode according to the light transmittance threshold. There are two working modes: one is the low-power mode, and the other is the high-performance mode. The low-power mode is when T ≥ 30%. The high-performance mode is when T < 30%.

[0060] When the low-power mode is working, the RISC-V main frequency drops to 100 MHz, and QPSK modulation is enabled. At this time, the symbol rate is 1 Msps, the transmit power ≤ 10 dBm, and the drone endurance time is extended from 90 minutes in the traditional solution to 120 minutes.

[0061] When the high-performance mode is working, the MMSE coprocessor and 16QAM modulation are activated. At this time, the symbol rate is 2Msps, the transmit power is increased to 20dBm, the LDPC code length is switched to 2048 bits, and the bit error rate is ≤10-6. At this time, the system starts multi-core parallel computing, and the task throughput is increased by 3 times.

[0062] Preferably, the dynamic resource management module also integrates an environmental temperature monitoring function. When the on-board temperature sensor detects that the chip temperature ≥ 75 °C, a frequency reduction strategy is automatically triggered (the main frequency is reduced from 1.5GHz to 800MHz), and the cooling fan is started (the rotation speed is 3000rpm) to ensure hardware stability. Tests show that this strategy reduces the peak chip temperature by 20 °C and avoids performance degradation caused by overheating.

[0063] The ground sensor node includes a temperature and humidity sensor, a soil moisture sensor, and a CO2 concentration detection module.

[0064] The temperature and humidity sensor uses an SHT35 chip, with a measurement range of -40 °C to 125 °C and an accuracy of ±0.5 °C.

[0065] The soil moisture sensor is based on the FDR principle, with a range of 0 to 100% Vol and a resolution of 0.1%. The model of the soil moisture sensor is Decagon EC-5.

[0066] The CO2 concentration detection module uses NDIR technology, with a range of 0 to 5000ppm and an error of ±50ppm. The model of the CO2 concentration detection module is Winsen MH-Z19.

[0067] The ground sensor node interacts with the drone through the NB-IoT protocol and uses the TDMA mechanism to transmit data in a time-sharing manner (the time slot interval is 10ms), thus avoiding channel conflicts. After receiving the data, the drone performs Huffman compression and AES-256 encryption through the RISC-V main control chip, and then forwards it to the command center through the satellite link. The actual measurement shows that the end-to-end transmission delay is ≤50ms and the packet loss rate is <0.1%.

[0068] Furthermore, the AES-256 encryption algorithm formula is where N is the amount of data to be encrypted (bytes), C AES is the encryption rate per second (bytes / second), cycles is the number of cycles for the RISC-V processor to process encryption, and T enc is the encryption time.

[0069] Furthermore, the system automatically adjusts the encryption strength according to the current communication environmental conditions. When the signal quality is poor, the system automatically increases the encryption strength to enhance the anti-interference ability of the data; while in the case of stable signals, the system reduces the encryption strength to reduce the computational overhead and energy consumption.

[0070] Preferably, the encryption strength formula is E adaptive = base_encryption × (1 + α × SNR current ), where base_encryption is the basic encryption strength, α is the encryption strength adjustment factor, and SNR current is the signal-to-noise ratio in the current environment, and E adaptive is the adaptive encryption strength.

[0071] In the preferred solution, the ground sensor node is built-in with a low-power wake-up circuit. When the UAV enters the communication radius (≤1 km), the node is woken up by the LoRaWAN beacon signal with a center frequency of 868 MHz, and automatically enters the sleep mode (power consumption ≤10 μA) after the data transmission is completed. The overall standby time of the node can reach 6 months, which is suitable for long-term unattended scenarios.

[0072] The satellite / base station communication link can adopt the following two solutions:

[0073] 1. Ka-band module: The operating frequency is 26.5 - 40 GHz, supporting a maximum transmission rate of 2 Gbps. The polarization multiplexing technology is adopted to improve the spectral efficiency, which is suitable for real-time backhaul of high-definition video streams (1080P@30fps).

[0074] 2. Starlink protocol device: Integrated with a 4×4 phased array antenna, achieving global coverage through a constellation of low-earth orbit satellites (orbital altitude 550 km), with a transmission delay ≤30 ms and a packet loss rate <0.01%.

[0075] Furthermore, the UAV monitors the signal strength (RSSI) of the satellite / base station communication link in real time during flight. When RSSI ≤ -90 dBm, it automatically switches to the standby frequency band or activates the relay mode (multi-hop transmission through neighboring UAVs) to ensure the reliability of the link. Tests show that in the extreme environment where the canopy transmittance <10%, the system can still maintain the communication link with a packet loss rate <5%.

[0076] The RISC-V chip can achieve task scheduling optimization through a customized operating system (such as FreeRTOS). The key algorithms adopted (such as beamforming, LDPC coding, etc.) are implemented in hardware description language (Verilog) and seamlessly integrated with the software layer API. Tests show that the hardware-software co-design improves the computing efficiency by 70% and reduces the power consumption by 35%.

[0077] Preferably, the system has a built-in self-check program to automatically check the hardware status (including antenna impedance matching, battery health status and memory remaining capacity) before flight. If an abnormality is detected (such as battery capacity <20%), an audible and visual alarm is triggered and takeoff is aborted, with a safety redundancy of 99.9%.

[0078] Furthermore, in order to cope with complex electromagnetic environments, the adaptive RF module integrates a frequency hopping sequence generator (frequency hopping rate 1000 times / second), supports dual-mode switching of pseudo-random sequence and adaptive sequence, and the interference avoidance success rate is ≥95%. The communication data packet adopts the AES-256 encryption algorithm, and the key update cycle is ≤10 minutes to prevent data theft and tampering.

[0079] In the preferred solution, the drone navigation system integrates GPS, Beidou and inertial navigation module (IMU, accuracy 0.1°). In signal-blocked environments (such as canyons or dense forests), the positioning error is ≤ 1 meter, and the path tracking accuracy is 60% higher than that of the pure GPS solution.

[0080] Example 2

[0081] This embodiment further describes the application of the forest canopy penetration communication relay UAV system based on RISC-V architecture in UAV-assisted edge computing, aiming to improve the real-time performance of data processing, reduce communication delays and optimize the overall performance of the UAV. By combining the computing power of the UAV platform and the edge computing architecture, the present invention can achieve efficient data processing and instant response in complex forest environments.

[0082] like Figure 1 As shown in the figure, the UAV system includes a UAV platform, ground sensor nodes, satellite / base station communication links, and a main control chip based on the RISC-V architecture. The UAV platform is equipped with a RISC-V main control chip with powerful computing power (such as the Pingtou Ge Yiying 1520 processor), which can support the execution of edge computing tasks. While performing communication relay tasks, the UAV can also perform real-time processing, analysis and storage of environmental data, reduce dependence on remote cloud servers, and reduce communication delays. The specific implementation process is as follows:

[0083] S1: The drone platform is equipped with a LiDAR module to collect real-time environmental data of the forest canopy during flight. At the same time, the ground sensor nodes also transmit data such as temperature and humidity, CO2 concentration, and soil moisture to the drone through the NB-IoT protocol. The drone's RISC-V main control chip receives this data and performs preliminary data processing and analysis through the embedded edge computing module.

[0084] S2: The main control chip of the drone processes data from sensors through a hardware acceleration module (such as an MMSE coprocessor) and a custom instruction set (such as the matrix operation instruction vbf.capon and the dynamic LDPC coding instruction crc.ldpc). The processed data includes the point cloud data from the LiDAR module, and based on this, a three-dimensional model of the forest canopy is built to calculate the light transmittance T, enabling dynamic switching of the working mode. At the same time, the data transmitted by the ground sensor nodes is compressed, filtered, and processed to extract key information (such as soil humidity changes, temperature changes, etc.), and predictive analysis is carried out through a custom algorithm model to provide real-time environmental monitoring reports.

[0085] S3: The results calculated by the drone are updated in real time to the internal database of the drone, and the communication and energy efficiency modes of the drone are adjusted according to the calculation results. For data with high real-time requirements, the RISC-V main control chip directly sends the results to the remote command center through a satellite link or a ground communication link; for data with higher tolerance to latency, the system stores the data in the on-board memory and uploads it in batches after the drone's flight mission is completed.

[0086] S4: The drone performs real-time evaluation of environmental conditions (such as forest canopy density, light intensity, humidity, etc.) through edge computing. Based on the evaluation results, the drone dynamically adjusts its working mode to optimize resource allocation. For example, in a sparse canopy environment, the system enables a low-power mode to reduce the consumption of computing resources; in a dense canopy environment, the system automatically switches to a high-performance mode to enhance computing power and communication stability.

[0087] S5: During the task execution, the drone continuously processes data and provides feedback. The edge computing module not only supports real-time analysis of data but also automatically adjusts the flight path according to environmental changes. After the task is completed, the drone returns autonomously through the A* path planning algorithm and confirms the hardware status through a self-check program to ensure the integrity and security of the task data.

[0088] Preferably, the edge computing module described in S1 is a computing module that completes data processing at the edge of the network, and the specific model is the Jetson ORIN NX 8GB core module. When working, the edge computing module receives data transmitted from the RISC-V main control chip and uses its high computing power to calculate the data, making up for the insufficient computing power.

[0089] Through the expansion of the RISC-V instruction set and hardware acceleration modules, the system of the present invention can achieve dynamic channel estimation, real-time beamforming, and adaptive LDPC coding, significantly improving the stability and real-time performance of the communication link. The present invention uses LiDAR technology to model the forest canopy transmittance in real time, dynamically switches between low-power and high-performance modes, optimizes the energy efficiency of the communication link, and maximizes the flight time of the drone. Through adaptive beamforming and biomimetic antenna design, the present invention effectively reduces the attenuation of signals by the forest canopy, improves the signal penetration ability, reduces the bit error rate and transmission delay. The present invention has efficient dynamic resource management capabilities, can operate stably in complex and dynamically changing forest environments, and is widely used in emergency communication scenarios such as forest fire warning, ecological environment monitoring, and post-disaster recovery. Compared with traditional communication relay solutions, the present invention has significant performance advantages and can effectively address various challenges in forest environments.

[0090] Compared with closed instruction set architectures such as ARM and x86, the open-source nature and modular design of the RISC-V instruction set adopted by the present invention provide significant customization advantages. The deep integration of its dedicated instruction extensions and hardware acceleration significantly reduces time delays and communication overhead and latency. At the same time, the minimalist instruction set and streamlined architecture of RISC-V not only improve real-time performance and computing efficiency but also achieve energy efficiency optimization and dynamic resource management. Moreover, the open-source ecosystem of RISC-V facilitates customization in vertical domains. Its characteristics of no license fees and hardware-software co-design greatly reduce costs and improve flexibility. Compared with other products using the RISC-V instruction set, the present invention mainly integrates the vbf.capon instruction and the crc.ldpc instruction, which are used for matrix operation optimization and dynamic coding acceleration respectively, to improve operation efficiency and reduce the bit error rate, thereby reducing CPU occupancy and enhancing adaptability to dynamic environments.

Claims

1. A forest canopy penetration communication relay UAV system based on the RISC-V architecture, characterized in that: It includes a drone platform, ground sensor nodes, and satellite / base station communication links; The drone platform is equipped with a RISC-V main control chip, an adaptive radio frequency communication module, and a bionic camouflage shell, which are used to realize the real-time establishment and dynamic adjustment of the forest canopy penetration communication link; The ground sensor nodes are used to collect forest land environment information and interact with the drone through a narrowband communication protocol; The satellite / base station communication link is used for data forwarding between the drone and the upper-level network.

2. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 1, wherein: The RISC-V main control chip adopts a 64-bit dual-core processor, with a built-in hardware acceleration module and a custom instruction set extension; among them, the hardware acceleration module supports custom vector and matrix processing extension instruction sets, The custom instruction set extension includes the vbf.capon instruction and the crc.ldpc instruction; the vbf.capon instruction is used to optimize the communication beam direction in real time, and the crc.ldpc instruction is used for channel adaptive modulation.

3. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 2, characterized in that: The vbf.capon instruction integrates a dedicated circuit for matrix inversion based on Cholesky decomposition, supporting real-time beamforming weight calculation; the formula for real-time beamforming weight calculation is W = R -1 a(θ); where R -1 is the covariance matrix, a(θ) is the beam direction angle, θ is the direction angle of beamforming, that is, the angle pointed by the main lobe of the beam; W represents the weight vector of beamforming.

4. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 2, wherein: The CRC.LDPC instruction dynamically adjusts the LDPC code length to match the optimal coding rate under the ITU-R F.1819 channel conditions in real time; the formula for dynamically adjusting the LDPC coding rate is k ∈ [256, 1536], n ∈ [512, 2048]; where: K is the information bit length, n is the total code length, and R is the LDPC coding rate.

5. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 2, characterized in that: The hardware acceleration module includes an MMSE co-processor and a butterfly bionic flexible antenna array; the MMSE co-processor is a channel estimation unit based on the minimum mean square error algorithm, integrating a channel matrix tracking and Doppler frequency shift compensation circuit; the butterfly bionic flexible antenna array adjusts the phase of the array antenna through a PWM signal controlled by a RISC-V processor.

6. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 1, characterized in that: The adaptive radio frequency communication module is designed based on the AD9361 chip, supports dynamic frequency hopping in the 700MHz - 6GHz wide frequency band, and has a spectrum sensing and frequency hopping mechanism.

7. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 1, characterized in that: The surface of the bionic camouflage shell outside the drone fuselage is covered with a chlorophyll spectral reflection coating.

8. A forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 1, characterized in that: A dynamic resource management module is set in the drone system, which combines the on-board LiDAR module to perform real-time modeling of the forest canopy light transmittance, and intelligently adjusts the processor frequency, modulation mode, and transmission power according to the model results.

9. The forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 8, wherein: The dynamic resource management module is a forest canopy density recognition mechanism based on lidar point clouds, which uses the light transmittance to judge the current channel environment and dynamically allocate system resources; when the light transmittance ≥ 30%, the drone system enables the low-power mode to ensure the basic reachability of communication; When the light transmittance < 30%, the drone system automatically activates the MMSE co-processor, the LDPC dynamic coding module, and the 16QAM modulation mode to maximize the performance of the communication link.

10. A forest canopy penetration communication relay UAV system based on the RISC-V architecture according to claim 9, characterized in that: The calculation formula for the light transmittance is where N ground is the number of laser points that penetrate the canopy and reach the ground, and N total is the total number of laser points emitted.

Images