Sushi-type S-shaped layered horizontal self-driven transmission mushroom and vegetable shelter planting system

Abstract

The invention relates to the technical field of general control or regulation systems, in particular to a sushi type S-shaped layered horizontal self-driven transmission mushroom and vegetable shelter planting system. The system comprises a data acquisition module, a path planning and scheduling module, a motor driving control module, an environment regulation and control module and a central controller. The system collects environment parameters, carrier states and crop growth information in real time through a sensor network, and constructs system state vectors. The central controller combines a heuristic path planning algorithm, an adaptive PID control algorithm, a model prediction control algorithm and a DQN optimization strategy based on deep reinforcement learning to realize carrier path optimization, motor driving precision control, cabin environment prediction regulation and control and system closed loop feedback optimization. The system has the advantages of flexible orbit scheduling, efficient operation response, fine environment control, high intelligent level and the like, and can realize subarea fine management and continuous optimization evolution of the system in a multi-category crop co-cabin planting scene.

Description

Technical Field

[0001] The present invention relates to the technical field of general control or regulation systems, and in particular to a sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system. Background Art

[0002] With the accelerating pace of urbanization and the rapid development of modern agriculture, integrated, intelligent, three-dimensional planting models are becoming increasingly important for improving agricultural production efficiency. With their advantages of small footprint, controlled environment, and flexible layout, container-type planting systems have been widely used in urban micro-farms, post-disaster agriculture, and specialized scientific research. Symbiotic cultivation of fungi and vegetables can, to a certain extent, achieve ecological advantages such as carbon dioxide-oxygen cycling and nutritional complementarity, offering high production efficiency and resource utilization potential.

[0003] However, current mainstream mushroom and vegetable cultivation systems still have many limitations: First, most systems use fixed or manually propelled cultivation racks and lack tracked vehicles and path planning capabilities, resulting in low utilization of planting space and inefficient operation scheduling. Second, environmental control is often based on simple closed-loop logic or static settings, making it difficult to meet the dynamic and precise requirements of different crops for light, humidity, and gas concentration. Furthermore, the systems generally lack intelligent learning and feedback mechanisms, making it impossible to achieve cross-module collaborative optimization and adaptive adjustment, which restricts the sustainable evolution of intelligent agricultural systems. Therefore, there is an urgent need for a composite mushroom and vegetable cultivation system with tracked scheduling, self-driven transmission, multivariable predictive control, and intelligent optimization capabilities to improve the automation, refinement, and intelligence of cultivation. Summary of the Invention

[0004] The purpose of the present invention is to provide a sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system to solve the problems of low space utilization, weak environmental coordination ability, low degree of automation and high energy consumption in the existing technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The mushroom and vegetable cubicle cultivation system of the present invention comprises:

[0007] The sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin cultivation system includes:

[0008] The data acquisition module collects environmental parameters, vehicle status and crop growth information in real time through the sensor network, and combines it with RFID to identify the crop growth stage, and uploads it to the central controller to construct the system state vector.

[0009] Path planning module: The central controller uses a heuristic algorithm to generate the optimal path based on track topology and target points, and dynamically corrects the path and timing with real-time feedback to achieve multi-vehicle scheduling and obstacle avoidance operation;

[0010] The drive control module, the central controller of which adjusts the frequency and start / stop state of the hysteresis linear motor based on the adaptive PID algorithm, realizes the precise propulsion, speed change and steering of the vehicle within the track;

[0011] The environmental control module uses a model predictive control algorithm to continuously predict environmental changes and optimize control strategies, dynamically adjusting parameters such as light, humidity, air pressure, and CO2 to meet crop zoning requirements;

[0012] Intelligent optimization module, the central controller has a built-in DQN reinforcement learning algorithm, which continuously updates the control strategy based on system feedback to achieve adaptive closed-loop optimization control of the path, motor and environment.

[0013] As a preferred embodiment of the present invention, the data collected by the data collection module includes:

[0014] Environmental parameters: temperature, humidity, CO2 concentration and light intensity;

[0015] Vehicle status: position, speed, motor frequency, current and power;

[0016] Crop information: Leaf area index, mycelium coverage image information, and growth stage recorded by RFID.

[0017] As a preferred embodiment of the present invention, the central controller includes a data receiving unit, a decision control unit and an execution coordination unit, including:

[0018] The data receiving unit is used to receive environmental parameters, vehicle status and crop information uploaded by the data acquisition module;

[0019] The decision control unit integrates path planning, adaptive PID, model predictive control and DQN optimization control algorithms to generate multi-objective control instructions;

[0020] The execution coordination unit is used to issue control instructions to the motor drive module and the environmental control module, and dynamically adjust the control strategy according to the feedback results to achieve multi-module collaborative closed-loop control.

[0021] As a preferred embodiment of the present invention, the track is a three-layer closed track structure, corresponding to the top vegetable area, the middle mushroom area and the bottom mushroom spare area respectively, and the corresponding layer spacing is preferably initially set to 60cm, 45cm and 30cm.

[0022] As a preferred embodiment of the present invention, the control process of the path planning module includes:

[0023] Construct the current track topology and vehicle occupancy status;

[0024] Analyze the target vehicle's RFID information and docking requirements;

[0025] Generate candidate paths based on heuristic search algorithm;

[0026] Combined with the path congestion scoring function, the optimal path is selected and control instructions are generated;

[0027] Send path control instructions to the motor drive module;

[0028] Modify paths and scheduling timing based on real-time feedback.

[0029] As a preferred embodiment of the present invention, the adaptive PID control process in the drive control module includes:

[0030] Collect the vehicle's position information, running speed and track curvature;

[0031] Calculate control deviation and dynamically adjust PID control proportional coefficient, integral coefficient and differential coefficient;

[0032] Generate motor control signals and send them to the target motor via the bus;

[0033] Receive motor feedback and update control parameters to form a closed-loop control.

[0034] As a preferred embodiment of the present invention, the model predictive control process of the environmental control module includes:

[0035] Establish a multivariate prediction model based on crop models and current environmental parameters;

[0036] Predict environmental change trends in the future;

[0037] Construct optimization objective function and generate control instructions;

[0038] Dynamically execute control instructions and update system status in real time to continuously optimize environmental factor settings.

[0039] As a preferred embodiment of the present invention, the DQN control strategy of the intelligent optimization module includes:

[0040] The central controller inputs the system state vector into the DQN neural network;

[0041] Output three types of control actions: path planning, motor drive and environment control;

[0042] Execute control actions and collect system responses to calculate reward values;

[0043] Generate state-action-reward triples and write them into the experience pool;

[0044] Periodically train and update DQN network weights;

[0045] The three types of control tasks are updated asynchronously to achieve closed-loop optimization control.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. Achieve efficient scheduling and precise obstacle avoidance for multi-vehicle track operations through a heuristic path planning algorithm. Compared to traditional fixed-path control methods, this invention dynamically generates optimal paths based on track topology, target location, and time window, effectively mitigating path conflicts and improving scheduling efficiency and space utilization.

[0048] 2. Adopting an adaptive PID control algorithm enables precise driving and dynamic steering within complex tracks. Compared to fixed-parameter PID control, this method can adjust control parameters in real time based on the operating status, improving motor response speed and operational smoothness, significantly enhancing system adaptability and robustness.

[0049] 3. Introducing a model predictive control algorithm to achieve rolling optimization and fine-tuning of multivariable environmental parameters. Compared to traditional fixed-value control methods, MPC uses crop models to predict future conditions and dynamically adjusts light, humidity, and gas concentrations, achieving both energy optimization and improved crop adaptability.

[0050] 4. Integrating deep reinforcement learning algorithms enables intelligent evolution and closed-loop optimization of path, motor, and environmental control strategies. Compared to manually set strategies, DQN can continuously self-learn and optimize control strategies based on system feedback, enhancing system intelligence, flexibility, and long-term operational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0052] In the attached figure:

[0053] Figure 1 This is a structural diagram of the mushroom and vegetable cubicle cultivation system based on the conveyor belt sushi-style S-shaped layered horizontal self-driven transmission of the present invention;

[0054] Figure 2 Schematic diagram of the S-type layered structure provided by the present invention;

[0055] Figure 3 This is a flow chart for realizing the mushroom and vegetable cubicle planting system based on the conveyor belt sushi-style S-shaped layered horizontal self-driven transmission of the present invention. DETAILED DESCRIPTION

[0056] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0057] Example 1:

[0058] In the sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system of the present invention, the sensor network is used to fully perceive the system operating environment and provide key input parameters to the central controller to realize data support for the control algorithm.

[0059] 1, including:

[0060] The data acquisition module collects environmental parameters, vehicle status and crop growth information in real time through the sensor network, and combines it with RFID to identify the crop growth stage, and uploads it to the central controller to construct the system state vector.

[0061] The data collected by the sensor network includes environmental parameters, vehicle status and crop growth information.

[0062] Specifically, the environmental data collection is performed by evenly distributing temperature and humidity sensors, CO2 concentration sensors, light sensors, and microclimate probe modules in different layers of the mushroom and vegetable cultivation cabin. Each type of sensor periodically samples and uploads the data to the edge data processing node via CAN bus or RS485 communication.

[0063] The vehicle data collection system integrates an inertial measurement unit, wheel encoders, and motor drive feedback modules at the bottom of each mobile vehicle carrying crops. The system records its current position information, operating speed, motor current, voltage, and power factor in real time. All status data is relayed to the central controller via Wi-Fi protocol.

[0064] Crop growth information collection involves installing a high-resolution image acquisition unit above each track, combining red and blue spectral analysis to perform image recognition and analysis of crop leaf area index and mycelium coverage. The recognition model is preferably deployed in a lightweight manner based on YOLOv5. Data processing is completed by edge devices, and feature results are uploaded to the control center. Each carrier is also attached with an RFID tag to identify the crop type, growth stage, sowing date, and management batch.

[0065] Furthermore, the system state vector is constructed as a central controller to perform timestamp synchronization, preprocessing and vector representation on the above-mentioned collected data. The construction of a unified system state vector will serve as the input basis of the reinforcement learning module to realize subsequent intelligent control decisions.

[0066] Path planning module: The central controller generates the optimal path based on the track topology and target point position using a dynamic path planning algorithm. It dynamically corrects the path and timing with real-time feedback to achieve multi-vehicle scheduling and obstacle avoidance operation.

[0067] The steps of the dynamic path planning algorithm include:

[0068] Construct the current track topology and load the occupancy status; parse the target vehicle's RFID identifier and target docking area; generate feasible paths based on a heuristic search algorithm; select the optimal path by combining the current time window and the path congestion prediction scoring function; output path instructions and transmit them to the hysteresis linear motor drive module; dynamically correct the path and timing based on the vehicle position feedback from the sensor.

[0069] Specifically, the track topology is modeled as a pre-set three-tier track system, corresponding to the top vegetable zone, the middle mushroom zone, and the bottom reserve zone. This structural information is stored in a central controller. The track is constructed as a directed graph structured by nodes and edges, with each node corresponding to a possible turning point, intersection, or docking area. Edge weights are initially set to the track segment length and can be dynamically adjusted based on vehicle density, curvature, and congestion during operation.

[0070] Furthermore, in the target mission analysis step, each vehicle uses RFID to identify its mission target and crop information. The central controller reads the RFID data to determine its next target docking area and track level, and then calibrates the coordinates of the path's starting and ending points.

[0071] The controller uses a heuristic search algorithm to generate multiple feasible paths and evaluates the paths based on the following multi-factor scoring function:

[0072] f(x)=g(n)+h(n)+ω1·C e +ω2·T e +ω3·K e

[0073] Where: g(n): the cumulative path distance of the current node n; h(n): the heuristic estimate from n to the target node; C e : The current congestion score of each track section in the path; T e : The average load of the path segment in the current time window; K e : Track curvature score, used to penalize paths with frequent turns; ω1, ω2, ω3: Score weights, adjustable by the system, with initial values ​​set to 0.4, 0.3, 0.3 by default.

[0074] After each path planning is completed, the system will save the top three feasible paths; for these candidate paths, the comprehensive scoring formula is as follows:

[0075]

[0076] Among them, α, β, and γ represent the degree of attention to speed, obstacle avoidance, and stability, respectively, and the initial values ​​are set to 0.5, 0.3, and 0.2. They represent the total execution time, the cross-conflict rate of other vehicles’ current paths, and the number of inter-layer transfers, respectively.

[0077] Finally, the path with the highest score is selected as the "main path" for this scheduling, and the remaining candidate paths are temporarily stored for backup.

[0078] If a path node n is detected to be occupied, a conflict warning is issued, or a blockage is detected during execution, the central controller immediately triggers local path replanning: the executed path segment is retained; a new path segment is generated using a lightweight heuristic search algorithm with the current position as the new starting point; and if the fault persists beyond the tolerance threshold, an alternative path is switched or a deceleration / wait instruction is issued to prevent system deadlock.

[0079] The drive control module, the central controller of which adjusts the frequency and start / stop state of the hysteresis linear motor based on the adaptive PID algorithm, realizes the precise propulsion, speed change and steering of the vehicle within the track;

[0080] The steps of adjusting each motor by the adaptive PID control algorithm include:

[0081] Collect the vehicle's current position, running speed, and track curvature information, calculate the deviation value, and input it into the PID controller; adaptively adjust the proportional coefficient K according to the deviation trend p , integral coefficient K i and differential coefficient K d , generate control signals to adjust the output frequency and current of the target motor; send them to the motor drive modules at each layer through the bus and execute them; collect feedback results after execution and correct the control parameters in a closed loop.

[0082] Specifically, the real-time data collection and input parameter construction steps are as follows:

[0083] The controller obtains the following key inputs from the sensor network in real time:

[0084] The current vehicle's position coordinates P t ; Travel speed V t ; The curvature radius R of the track segment t or path curvature κ t ; The current operating status of the motor, including frequency, current, and power.

[0085] Construct the input deviation of the PID controller:

[0086] ΔP=P target -P t

[0087] ΔV=V target -V t

[0088] Where ΔP: position deviation, P target : Target path point, ΔV: Speed ​​deviation, V target : expected speed.

[0089] Furthermore, the dynamic adjustment of K p , K i , K d Parameter steps:

[0090] The central controller dynamically adjusts the PID parameters based on the current load status and track complexity using empirical rules and a fuzzy logic rule base:

[0091] K p : Increase K in the straight line segment p To improve the response speed; reduce K in the corners p To avoid overshoot; K i :When the deviation lasts for a long time but the amplitude is small, the system automatically increases K i To eliminate steady-state error; K d : Increase K when the track curvature is large or the speed changes rapidly d , to smooth the control curve.

[0092] If ΔP continues to be too large and ΔP changes dramatically, increase K p , K d If ΔP is stable but slowly approaches the target, increase K i ; If the system oscillates or overshoots, reduce K p , K d .

[0093] Furthermore, the control signal is generated and sent as follows:

[0094] According to the adjusted K p , K i , K d Coefficient, use the PID formula to calculate the target control output U(t):

[0095]

[0096] Where e(t) is the current comprehensive deviation, that is, the weighted fusion of position and velocity.

[0097] The control signal is converted into the target current frequency value and start / stop status signal by the central controller and sent to the motor drive module.

[0098] Each layer of hysteresis linear motor receives central instructions through a dedicated control node to achieve: start / stop logic control; speed regulation; fine position propulsion.

[0099] The controller schedules multiple motors in parallel through the CAN bus to achieve synchronous propulsion and staggered control of the vehicle on each track layer to avoid interference and collision.

[0100] The controller collects the execution results in real time and compares them with the expected values. If there is overshoot, steady-state error or slow response, the error behavior is automatically recorded. The error signal is fed back to the control logic module to p , K i , K d Perform adaptive fine-tuning corrections; at the same time, record them in the control log as reinforcement learning experience data input for the DQN optimization module.

[0101] The environmental control module uses a model predictive control algorithm to continuously predict environmental changes and optimize control strategies, dynamically adjusting parameters such as light, humidity, air pressure, and CO2 to meet crop stratification requirements;

[0102] The model predictive control algorithm implements the refined control steps including:

[0103] A multivariable state prediction model is established based on the crop growth model, and the environmental change trend in future time steps is predicted based on the acquired real-time environmental parameters; an objective function is constructed to minimize the error between the predicted value and the set environmental parameters, and the optimization problem is solved to generate control instructions; the control instructions are executed and the system status is updated in real time; the above prediction and control process is executed cyclically to continuously adjust the environmental control parameters.

[0104] Specifically, the steps for building a multivariable state prediction model are as follows: the system first accesses real-time environmental data collected by the sensor network, including temperature, humidity, carbon dioxide concentration, light intensity, and air pressure;

[0105] At the same time, combining the crop growth stage model with historical crop growth data, we can establish a mapping table of optimal environmental parameters corresponding to the growth status of crops at different levels.

[0106] Use the above data to build a multi-input multi-output state space model:

[0107] X t+1 =AX t +BU t +W t

[0108] Y t =CX t +DU t +V t

[0109] Where X represents the system state vector (including environmental factors), U represents the control input, Y is the output observation, and W t and V t are process and observation noise, A, B, C, and D respectively represent the system state transfer matrix, the mapping matrix of the influence of control input on system state, the mapping matrix from state vector to observable output, and the mapping matrix of control input directly acting on observation output.

[0110] Set the rolling prediction horizon to N steps;

[0111] At the current time t, based on the system state X t and the control variable U t , use the model to recursively infer the state value of each step in the future {X t+1 ,X t+2 ,...,X t+N};

[0112] At the same time, the output value {Y t+1 ,Y t+2 ,...,Y t+N}, used to assess the medium- and short-term impacts of environmental factors on crop growth.

[0113] Furthermore, the optimizing function and generating the control instruction include:

[0114] Construct the optimization objective function J, the form of the optimization objective function is:

[0115]

[0116] Where Q and R are weight matrices, ΔU is the control input change, which is obtained by taking the difference between the current control input vector and the control input vector of the previous control cycle, and Y * is the target value of the system output, and T is the transpose symbol of the matrix.

[0117] Use the quadratic programming solver to solve the optimization problem and obtain the optimal control input sequence {U t ,U t+1 ,...,U t+N-1};

[0118] Only the first control action U is implemented t , as part of the "rolling horizon" mechanism;

[0119] Control instructions are sent to the environmental execution subsystem via the communication bus to perform the following operations on various devices: controlling the intensity and lighting cycle of the LED lighting panel; controlling the operating time of the atomizer and the output frequency of the water pump; adjusting the wind speed and direction of the ventilation system to achieve two-way control of humidity and CO2 concentration; and adapting to different crop types in the cabin to achieve layered and zoned environmental parameter configuration.

[0120] After each control action is executed, the system re-collects the current environmental status and crop feedback data;

[0121] Based on the new state update system model, the steps are repeated to roll out the prediction of future state sequences and solve and issue control instructions, thus realizing a real-time rolling prediction and dynamic control closed-loop mechanism;

[0122] The system also records the control effects and energy consumption data of each round, providing training samples for model fine-tuning and reinforcement learning modules during long-term operation.

[0123] Intelligent optimization module, the central controller has a built-in DQN reinforcement learning algorithm, which continuously updates the control strategy based on system feedback to achieve adaptive closed-loop optimization control of the path, motor and environment.

[0124] The central controller asynchronous adaptive optimization control step includes:

[0125] The central controller receives real-time data from the sensor network and constructs a system state vector. The state vector includes the vehicle's current position, speed, current motor frequency, temperature and humidity, CO2 concentration, light intensity, and image recognition parameters. The state vector is input into a pre-trained DQN neural network, which outputs optimal action instructions for three types of operations: vehicle movement, motor control, and environmental regulation. The action instructions are executed to control the vehicle's start and stop, motor parameter adjustment, and real-time setting of environmental factors. The central controller evaluates the immediate feedback of each set of actions and generates a reward value, forming a state-action-reward triple. The triple data is stored in the experience replay pool and the DQN network parameters are periodically updated. The three types of control objects adopt an asynchronous policy update mechanism, and the central controller generates control instructions according to the module's required frequency to achieve modular closed-loop optimization control of the system operation.

[0126] Specifically, the central controller obtains the following data dimensions from the sensor network and the visual acquisition unit in each cycle to construct a complete system state vector S t , mainly including:

[0127] Vehicle status parameters: current level, position coordinates, running speed, motor drive frequency and output current;

[0128] Environmental parameters: temperature, humidity, CO2 concentration, and light intensity of each layer;

[0129] Crop image features: visual recognition results such as leaf area index and mycelium coverage;

[0130] RFID identification information: the growth stage and regional location information of the crops currently carried by the vehicle;

[0131] Scheduling background information: track topology, path usage density, load distribution, etc.

[0132] Furthermore, the above state vector S t Input is fed into a pre-trained DQN deep neural network, which is locally deployed by the central controller and trained for the task through multiple rounds of simulation reinforcement training;

[0133] The network outputs the Q value set Q(S t ,a), respectively corresponding to the expected reward value of each candidate action in the three types of action spaces: path scheduling, motor driving, and environment control;

[0134] For each type of control task, select the action with the largest current Q value from the corresponding action set:

[0135] Path control: select new trajectory readjustment instructions or maintain the current path;

[0136] Motor control: select acceleration / deceleration, start / stop or adjust frequency strategy;

[0137] Environmental Control: Select the temperature, humidity, light, and gas concentration settings that are appropriate for the current forecast environment.

[0138] The system sends the above three types of optimal action instructions to the corresponding execution modules through the bus;

[0139] After the control action is executed, the system delays for one cycle to collect feedback results, including: the changing trend of the crop image; the motor response effect and environmental control effect recorded by the sensor; and the deviation between the control target and the actual feedback value.

[0140] Furthermore, the instant reward value calculation and experience triplet generation are performed by the central controller based on the above feedback to calculate the instant reward value R for each type of control task. t , the design is as follows:

[0141] Path control rewards: obstacle avoidance success rate, whether driving time is optimized, and whether the target area is reached;

[0142] Motor control rewards: driving stability, energy consumption indicators, and whether the response time is shortened;

[0143] Environmental control rewards: reduction in environmental deviation, energy consumption per unit time, and improvement rate of crop indicators;

[0144] And generate the corresponding experience triples (S t ,A t ,R t ,S t+1 ), which correspond to the current state; the three types of control actions taken; the immediate reward; and the state at the next moment.

[0145] All experience triplets are written into three independent experience replay pools, corresponding to the three types of control objects: path, motor, and environment;

[0146] The central controller randomly samples a number of triplets from each experience pool in proportion every fixed training cycle to update the network parameters;

[0147] Backpropagation training of the Q network is performed using the mean square error loss function:

[0148]

[0149] in: a means in state S t+1 Among all possible actions a, the one with the largest Q value is selected. It reflects the agent's estimate of the "optimal decision in the future" and is a key part of DQN update. γ is the discount factor.

[0150] The DQN modules of various control objects are updated separately in an asynchronous manner to avoid interference and support the modular evolution of the system.

[0151] Furthermore, the system supports strategy evolution and behavior migration during long-term operation. Through continuously accumulated experience samples and periodic update mechanisms, the model can be migrated from static optimality to dynamic adaptive optimality.

[0152] The central controller automatically tracks the control effects of different modules and determines whether the learning frequency of a sub-module needs to be increased to solve the control mismatch problem caused by environmental changes, load changes or crop distribution changes;

[0153] A cross-module collaborative closed-loop control mechanism for cluster vehicle scheduling, environmental collaborative control and drive strategy optimization has been implemented, improving the overall intelligence level of the system.

[0154] Example 2:

[0155] This embodiment is based on Figure 2 The structure of the mushroom and vegetable cabin shown in the figure further illustrates the specific deployment and coordinated operation mode of the system of the present invention in a closed integrated space, so as to realize the whole process of mushroom and vegetable symbiotic cultivation, precise control and intelligent scheduling under multi-layer tracks, and execute the following Figure 3 method flow.

[0156] The S-shaped layered structure, the mushroom and vegetable cabin includes a three-layer S-shaped closed ring track group.

[0157] Specifically, such as Figure 1As shown, the mushroom and vegetable cabin in this embodiment has a size of 6m×3m×2.5m, and is equipped with three layers of S-shaped closed tracks inside. Each layer of track is arranged in a revolving sushi-style layout along the horizontal plane of the cabin. The tracks are connected by a vertical revolving structure to form a continuous closed-loop operation path, realizing the circulation of cultivation vehicles. The inter-layer spacing of the three-layer track is 60cm, 45cm and 30cm respectively. In order to meet the growth needs of different crops, a hydraulic rod mechanism is integrated on the track column to support stepless adjustment of the track height within the range of 0.3 meters to 1.2 meters, thereby adapting to the differentiated requirements of different plants for light space and vertical environment.

[0158] The horizontal self-drive transmission device uses an adaptive PID control algorithm to adjust the driving frequency and start and stop status of each motor through the central controller to achieve precise propulsion and steering of the vehicle within the track;

[0159] Specifically, the drive structure incorporates a hysteresis linear motor stator embedded within each track layer, and a corresponding rotor structure embedded within the bottom of the cultivation vehicle, forming a complete magnetic suspension drive pair and enabling contactless horizontal self-propelled propulsion. A single vehicle has a maximum load of 50 kg and a rated operating speed of 0.2 m / min. It also features precise start-stop, fixed-point steering, and path correction capabilities.

[0160] To improve the system's energy efficiency and operational independence, a photovoltaic coating is added to the bottom of each cultivation vehicle, which together with the solar panels installed on the top of the cabin forms a complementary microgrid to ensure low-power continuous operation of the sensors and control modules in the vehicle.

[0161] The environmental control subsystem dynamically adjusts the operating parameters of environmental factors based on the data collected by the sensor network through the central controller based on the model predictive control algorithm, thereby achieving refined management of the hierarchical environment of fungi and vegetables;

[0162] Specifically, in terms of environmental perception, multiple groups of sensor nodes are evenly distributed inside the cabin, including SHT35 temperature and humidity sensors and MH-Z19B carbon dioxide sensors, covering all cultivation track partitions, and uploading the collected data to the central controller in real time through wireless communication.

[0163] To achieve zoning control of the mushroom-vegetable symbiosis, an adjustable LED lighting array is installed on the top of the cabin, and combined with interlayer isolation membranes and micro wind pressure adjustment mechanisms, three independent environmental zones are formed:

[0164] The top area is a high-light vegetable area, with light intensity maintained at 15,000 lux, which is used for the rapid growth of vegetable crops such as spinach; the middle track is a low-light mushroom area, with light intensity maintained at 200 lux, which is suitable for the cultivation of mushrooms such as shiitake mushrooms and enoki mushrooms; the bottom area is a dormant area with the light source turned off, which is used for empty vehicles to dock or temporarily store unharvested cultivated bodies.

[0165] The central controller is communicatively connected to the hysteresis linear motor system, the environmental control subsystem and the sensor network. The central controller has a built-in DQN-based reinforcement learning algorithm. The algorithm controls the vehicle path, motor drive parameters and environmental operating status based on the data collected by the sensor network to achieve asynchronous adaptive optimization control.

[0166] Specifically, the central control system uses an STM32 embedded controller as the core computing unit, integrating a multi-sensor data fusion module and a DQN-based reinforcement learning algorithm. The system accurately identifies crop type, growth stage, and location by reading information from RFID tags integrated into the vehicle, and schedules the movement of vehicles between different tracks based on a dynamic path planning algorithm:

[0167] In the empty state, the vehicle first navigates to the mushroom inoculation area and completes the loading of mushroom sticks or seeds;

[0168] The scheduling algorithm then guides it into the middle orbit to carry out the main fungus cultivation cycle (set to 8 days);

[0169] After the cultivation period, the system dispatches the robotic arm to transport the fungus culture to the top vegetable area, entering the fungus-vegetable symbiotic stage;

[0170] The system continuously tracks the cultivation progress and determines the maturity status by combining image recognition and historical data.

[0171] After the harvesting decision is made, the vehicle automatically drives to the exit, where manual or automatic harvesting equipment performs terminal processing.

[0172] Furthermore, the system integrates an adaptive PID control algorithm to adjust motor operating parameters. Based on the vehicle's real-time position, track curvature, and current speed, it dynamically calculates and corrects output values, improving track stability and accuracy. Simultaneously, a model predictive control algorithm is introduced to provide feedforward prediction and adjustment of the temperature, humidity, and light environment. Through real-time environmental perception and crop model simulation, this algorithm minimizes expected errors and optimizes resource allocation.

[0173] Finally, the system synergistically integrates the above modules into the DQN reinforcement learning control framework, optimizes the environmental control, path scheduling and drive execution links respectively, and forms a closed adaptive control loop to ensure that the system can achieve optimal performance under different load, environmental and task conditions.

[0174] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. Sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system, characterized by: include: The data acquisition module collects environmental parameters, vehicle status, and crop growth information in real time through a sensor network, and uses RFID to identify the crop growth stage, and uploads the data to the central controller to construct a system state vector. Path planning module: The central controller uses a heuristic algorithm to generate the optimal path based on track topology and target points, and dynamically corrects the path and timing with real-time feedback to achieve multi-vehicle scheduling and obstacle avoidance operation; The drive control module, the central controller of which adjusts the frequency and start / stop state of the hysteresis linear motor based on the adaptive PID algorithm, realizes the precise propulsion, speed change and steering of the vehicle within the track; The environmental control module uses a model predictive control algorithm to continuously predict environmental changes and optimize control strategies, dynamically adjusting light, humidity, air pressure, and CO2 concentration to meet crop stratification requirements; Intelligent optimization module, the central controller has a built-in DQN reinforcement learning algorithm, which continuously updates the control strategy based on system feedback to achieve adaptive closed-loop optimization control of the path, motor and environment.

2. The mushroom and vegetable shelter planting system according to claim 1 is characterized in that: The data collected by the data collection module includes: Environmental parameters: temperature, humidity, CO2 concentration and light intensity; Vehicle status: position, speed, motor frequency, current and power; Crop information: Leaf area index, mycelium coverage image information, and growth stage recorded by RFID.

3. The mushroom and vegetable shelter planting system according to claim 1, characterized in that: The central controller includes a data receiving unit, a decision control unit and an execution coordination unit, wherein: The data receiving unit is used to receive environmental parameters, vehicle status and crop information uploaded by the data acquisition module; The decision control unit integrates path planning, adaptive PID, model predictive control and DQN optimization control algorithms to generate multi-objective control instructions; The execution coordination unit is used to issue control instructions to the motor drive module and the environmental control module, and dynamically adjust the control strategy according to the feedback results to achieve multi-module collaborative closed-loop control.

4. The sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system according to claim 1 is characterized in that: The track is a three-layer closed track structure, corresponding to the top vegetable area, the middle mushroom area and the bottom mushroom spare area respectively, and the corresponding layer spacing is initially set to 60cm, 45cm and 30cm.

5. The sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system according to claim 1 is characterized in that: The control process of the path planning module includes: Construct the current track topology and vehicle occupancy status; Analyze the target vehicle's RFID information and docking requirements; Generate candidate paths based on heuristic search algorithm; Combined with the path congestion scoring function, the optimal path is selected and control instructions are generated; Send path control instructions to the motor drive module; Modify paths and scheduling timing based on real-time feedback.

6. The sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system according to claim 1 is characterized in that: The adaptive PID control process in the drive control module includes: Collect the vehicle's position information, running speed and track curvature; Calculate the control deviation and dynamically adjust the PID control proportional coefficient K p , integral coefficient K i and differential coefficient K d ; Generate motor control signals and send them to the target motor via the bus; Receive motor feedback and update control parameters to form a closed-loop control.

7. The sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system according to claim 1 is characterized in that: The model predictive control process of the environmental control module includes: Establish a multivariate prediction model based on crop models and current environmental parameters; Predict environmental change trends in the future; Construct optimization objective function and generate control instructions; Dynamically execute control instructions and update system status in real time to continuously optimize environmental factor settings.

8. The sushi-style S-shaped layered horizontal self-driven mushroom and vegetable cabin planting system according to claim 1 is characterized in that: The DQN control strategy of the intelligent optimization module includes: The central controller inputs the system state vector into the DQN neural network; Output three types of control actions: path planning, motor drive and environment control; Execute control actions and collect system responses to calculate reward values; Generate state-action-reward triples and write them into the experience pool; Periodically train and update DQN network weights; The three types of control tasks are updated asynchronously to achieve closed-loop optimization control.

Images