Robot delivery planning method based on full life cycle carbon emission evaluation and optimization

Abstract

The invention provides a robot delivery planning method based on full-life-cycle carbon emission evaluation and optimization. The robot delivery planning method comprises the following steps: S1, constructing a full-life-cycle total carbon emission model; s2, constructing a multi-objective optimization model: defining decision variables, constraint conditions and a dual-objective optimization model; s3, dynamically adjusting the weight of the multi-model optimization model, then solving the multi-target optimization model by adopting an NSGA-II algorithm, and obtaining a Pareto optimal solution set for a user to select a Pareto optimal solution therein; and S4, based on the selected robot deployment scheme, correcting the Pareto optimal solution set. According to the invention, economic delivery optimization of the robot is realized, and intelligent decision support is provided for low-carbon economic transformation.

Description

Technical Field

[0001] This invention belongs to the field of robot application and resource optimization technology, specifically involving a robot deployment planning method based on full life cycle carbon emission assessment and optimization. Background Technology

[0002] With the rapid development of industrial automation and intelligent manufacturing, robots are widely used in logistics, manufacturing, and service industries. While their large-scale deployment brings significant economic benefits, it also triggers a surge in carbon emissions throughout their entire lifecycle. Traditional robot deployment solutions often focus on functional implementation, operational efficiency, and cost control, lacking a systematic assessment and optimization of environmental impact, especially carbon emissions. This leads to the robot industry facing the challenge of sustainable development characterized by high energy consumption and high carbon emissions.

[0003] In existing technologies, improvements to the environmental impact of robots mainly focus on local aspects: Patent CN110850789A proposes a robot energy efficiency optimization method based on edge computing, but it only focuses on the usage stage and does not cover other high-carbon emission stages such as manufacturing and transportation; Patent CN112729307A proposes a robot path planning method based on multi-sensor data, but does not include carbon emissions in the objective function.

[0004] In summary, the existing technology has the following shortcomings: Carbon emission accounting is one-sided: it is limited to a single stage and lacks a quantitative model for carbon emissions that covers the entire life cycle of "resource extraction-production-transportation-deployment-operation and maintenance-disposal"; lack of multi-objective coordination: it is difficult to dynamically balance economic costs and carbon emission targets, which limits the practicality of the solution; insufficient dynamic optimization capability: it fails to dynamically adjust the investment strategy in conjunction with changes in energy structure and carbon tax policies, making it difficult to adapt to the needs of low-carbon development. Summary of the Invention

[0005] The purpose of this invention is to provide a robot deployment planning method based on full life-cycle carbon emission assessment and optimization, achieving economical robot deployment optimization and providing intelligent decision support for low-carbon economic transformation. The technical solution adopted is as follows: A robot deployment planning method based on full life-cycle carbon emission assessment and optimization includes the following steps: Step S1: Construct a full life-cycle carbon emission total model; Step S2: Construct a multi-objective optimization model: Define decision variables, constraints, and a bi-objective optimization model. ; Step S3: Dynamically adjust the multi-model optimization model The weights are then determined, and the NSGA-II algorithm is used to solve the multi-objective optimization model. Obtain the Pareto optimal solution set for users to select from the Pareto optimal solutions; Step S4: Based on the selected robot deployment scheme, correct the Pareto optimal solution set.

[0006] Preferably, step S2 specifically includes the following steps: Step S2A: Define decision variables ; Step S2B: Establish a system based on total carbon emissions throughout the entire life cycle. and based on total cost Bi-objective optimization model : Step S2C: Set constraints.

[0007] Preferably, decision variables for: ; in, - Number of deployments; - Energy type Related robot operation paths; - Maintenance strategy functions related to time 𝜏.

[0008] Preferably, step S3 specifically includes the following steps: Step S3A: Introduce the entropy weight method to dynamically adjust the weights of economic objectives. Weighting of environmental protection goals ; Step S3B: Solve for the Pareto optimal solution set using the NSGA-II algorithm; Step S3C, user selects Pareto optimal solution.

[0009] Preferably, step S4 specifically includes the following steps: Step S4A: Real-time acquisition of actual operating data in the virtual environment; Step S4B: Update the parameters of the NSGA-II algorithm online using the DQN algorithm, and then execute step 3 to obtain the corrected Pareto optimal solution set.

[0010] Preferably, the reward function of the DQN algorithm is: ; in, and This is a historical benchmark value. , This is the reward coefficient; - Total carbon emissions data from actual operation; - Total cost data from actual operating data.

[0011] Compared with the prior art, the advantages of the present invention are: A carbon accounting mechanism that integrates the entire lifecycle with blockchain: By dividing the data into stages and storing it on the blockchain, it solves the problems of data silos and low credibility in traditional methods.

[0012] Dynamic weighted multi-objective optimization: Combining the entropy weight method with the NSGA-II algorithm to achieve an adaptive balance between environmental protection and economic goals.

[0013] Digital twin-driven closed-loop optimization: Using DQN reinforcement learning to feed back virtual verification results to the actual system (an integrated hardware platform consisting of data acquisition modules, computing core modules, interactive terminals, and blockchain nodes), improving the reliability of the solution and reducing trial and error costs. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the operational process of a robot deployment planning method based on full life-cycle carbon emission assessment and optimization. Figure 2 A diagram illustrating the division of carbon emissions throughout the entire life cycle; Figure 3 Chart showing the contribution of robots to carbon emissions throughout their entire lifecycle; Figure 4 A traceability diagram of the robot's production process; Figure 5 A flowchart illustrating the closed-loop process for multi-objective optimization and digital twin verification; Figure 6 This is a detailed roadmap of the overall technical architecture for a robot deployment planning method based on full life-cycle carbon emission assessment and optimization. Detailed Implementation

[0015] The robot deployment planning method based on full life-cycle carbon emission assessment and optimization of the present invention will be described in more detail below with reference to the schematic diagrams, which illustrate preferred embodiments of the present invention. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the present invention.

[0016] like Figures 1-6 A robot deployment planning method based on full life-cycle carbon emission assessment and optimization includes the following steps: Step S1: Construct a full life-cycle carbon emission model .

[0017] ; - Direct carbon emissions, i.e. carbon emissions generated by fuel combustion or electricity consumption during robot operation, based on real-time grid carbon emission factors (e.g., the average factor for the Chinese regional grid is 0.55 kgCO2 / kWh). - Implicit carbon emissions, i.e. indirect emissions from material production, transportation, etc., are matched through a dynamic factor library.

[0018] Calculation in step S3B Direct carbon emissions are dynamically calculated based on real-time grid carbon emission factors, while implicit carbon emissions are calculated by matching indirect emission data from material production, transportation, etc., with a dynamic factor library. The robot's lifecycle is divided and data is collected to build a carbon emission calculation model and trace the blockchain data.

[0019] Among them, such as Figure 2 As shown, the robot lifecycle is divided into five stages: resource extraction (A), production and manufacturing (B), transportation and deployment (C), operation and maintenance (D), and end-of-life recycling (E). The following data is collected for each stage: Resource extraction (A): Carbon footprint of raw materials (such as metals and plastics) (unit: kgCO2 / kg), obtained through international standards databases; Manufacturing (B): Production energy consumption (electricity, fuel) and process carbon emissions (such as welding, spraying) are monitored in real time through factory production data and sensors; Transportation deployment (C): Transportation distance, vehicle type (air / land / sea) and corresponding carbon emission factor (kgCO2 / km·t); Operation and maintenance (D): Implicit carbon in robot operation energy consumption (positively correlated with task load and path length) and maintenance frequency (such as battery replacement and parts repair); End-of-life recycling (E): recovery rate (metal material recovery rate >90%), carbon emission reduction benefits of recycled materials (the carbon footprint of recycled aluminum is 5% of that of primary aluminum).

[0020] Blockchain technology is used to store carbon data throughout the entire process. Information on raw material suppliers, logistics records and energy consumption data are collected at each stage. A unique hash value is generated and stored on the blockchain. The integrity and consistency of the data are automatically verified through smart contracts to ensure that carbon accounting is transparent and traceable.

[0021] Direct carbon emissions = Energy consumption × Real-time grid carbon emission factor Implicit carbon emissions = Σ (material quantity × material carbon footprint) + Σ (transport distance × transportation carbon emission factor) + Σ (energy consumption × corresponding emission factor) + others In one specific embodiment, the carbon footprint of lightweight fuselage materials (such as carbon fiber) is 25 kgCO2 / kg; the carbon footprint of lithium battery production is 80 kgCO2 / kWh.

[0022] In a specific embodiment, such as Figure 3 As shown in the figure, carbon dioxide is the primary source of greenhouse gas emissions during the robot's entire lifecycle, contributing significantly to total carbon emissions. While methane and other greenhouse gases have a strong greenhouse effect, their emissions are relatively low. This finding is beneficial for guiding production, transportation, operation, maintenance, and recycling processes, and allows for the development of targeted emission reduction strategies for specific greenhouse gases.

[0023] In a specific embodiment, such as Figure 4 As shown, blockchain data should be traced back to its source.

[0024] This diagram can be used to analyze and manage the different sources of materials and components in the supply chain, helping companies optimize their procurement and production processes; It can also help assess the environmental impact of the supply chain, particularly in terms of carbon emissions and resource consumption, by identifying different material markets; it also helps to understand the distribution and price fluctuations of different material markets, which is helpful for cost control and budget planning.

[0025] Blockchain technology is used to reliably store carbon emission data. Carbon data at each stage (such as raw material supplier information and transportation logistics records) generates hash values ​​and is uploaded to the chain. Data consistency is automatically verified through smart contracts to ensure the transparency and traceability of carbon emission accounting.

[0026] Step S2: Construct a multi-objective optimization model.

[0027] Specifically define the decision variables and constraints, and design a bi-objective optimization model.

[0028] Step S2A: Define decision variables .

[0029] Number of deployments ( ): The number of robots in the deployment area, which is an integer variable; Path planning ( ): with energy type Related robot operation paths; Energy consumption and task duration are affected by continuous variables and are influenced by energy type. Battery life constraints; Energy type combination ( Energy types include grid-connected, fuel cell, or hybrid energy sources. For discrete variables; Maintenance strategy function ( Preventive maintenance cycle, It affects the frequency of parts replacement and the amount of hidden carbon emissions; It is a periodic decision-making process, and is time-dependent.

[0030] Right now ; in, = ; = ； {Power grid, fuel cell, hybrid mode}; A code indicating the type of energy; Step S2B: Establish a system based on total carbon emissions throughout the entire life cycle. and based on total cost Bi-objective optimization model : Dual-objective optimization model: ; ; = Purchase cost + Energy consumption cost + Maintenance cost + Carbon tax cost; - Total carbon emissions throughout the entire life cycle; - Covers robot purchase, energy consumption, maintenance costs, and carbon tax; Step S2C: Set constraints.

[0031] Ensure that the task completion time is within a reasonable range and that the robot's load capacity is in line with the task requirements.

[0032] (1) Task completion constraints: ; in, - Robot A's effective payload speed at time A; -Task assignment flag function; - Overall task requirements.

[0033] Robot payload speed: This refers to the robot's effective working speed at time t while performing a task, obtained from the real-time data interface of the robot controller or task management system. This value is typically determined by the robot model, payload weight, and the current task type.

[0034] Task assignment flag function: This is a binary variable used to indicate whether robot r has been assigned a task at time t. = 1 indicates that the robot is currently executing a task. = 0 indicates that it is in an idle, charging, or moving state.

[0035] Total task requirements: refers to the total amount of work that needs to be completed, and is used as an input parameter.

[0036] (2) Energy supply constraints: ; in, - Energy type At any moment Power requirements; -Energy availability factor; - Energy capacity limit.

[0037] Energy type At any moment Power demand: refers to the power requirement of energy type at time t. The power requirements are based on real-time energy consumption monitoring data during robot operation.

[0038] Energy availability coefficient; a coefficient between 0 and 1, representing the availability of energy type e_k at time t. Obtained from BMS.

[0039] (3) Dynamic coupling constraints on carbon emissions: ; in, -Real-time energy consumption rate of robot 𝑟 at stage 𝑖; -Dynamic function of power grid carbon emission factor; - Rate of change in implicit carbon stock; - Material carbon conversion rate.

[0040] Dynamic function of power grid carbon emission factor: refers to the carbon emissions corresponding to each kilowatt-hour of electricity consumed by the regional power grid at time t. It is obtained from official real-time data interfaces or dynamic databases released by national or regional energy management agencies.

[0041] Implicit carbon stock change rate: refers to the rate of change of implicit carbon emissions stock due to material consumption or replenishment in stage i.

[0042] It is calculated using blockchain-based notarized bill of materials (BOM) and logistics data.

[0043] Material carbon conversion rate: This represents the carbon emissions generated by acquiring or processing a unit of material in stage i. It is obtained by matching data from the international standard Life Cycle Assessment (LCA) database or the company's own dynamic factor library.

[0044] (4) Maintenance strategy nonlinear constraints: ; Indicates maintenance costs vary with maintenance cycle It exhibits a mixed exponential-logarithmic relationship; , , This is the attenuation coefficient for physical robot equipment.

[0045] Physical equipment attenuation coefficient: A physical coefficient that characterizes the performance degradation pattern of a specific robot model or its components. It is derived through big data analysis and machine learning fitting of historical operation and maintenance data of the same model of equipment.

[0046] This constraint is a specific component of F2, characterized by long maintenance cycles and high preventative maintenance costs. The formula will increase, the maintenance cycle will be shorter, and the exponential formula will increase, which means that the maintenance cost after the failure will increase. Therefore, this formula is also used as a constraint.

[0047] Step S3: Dynamically adjust the multi-model optimization model The weights are then determined, and the NSGA-II algorithm is used to solve the multi-objective optimization model. This allows you to obtain a set of Pareto optimal solutions, which users can then select from.

[0048] Step S3A: Introduce the entropy weight method to dynamically adjust the weights of economic objectives. Weighting of environmental protection goals .

[0049] ; ; in, -Real-time carbon tax price (RMB / ton CO2); -Energy market price volatility coefficient; , These are all user preference parameters (configurable via the interactive interface).

[0050] Step S3B: Use the NSGA-II algorithm to solve for the Pareto optimal solution set.

[0051] The Pareto optimal solution set is the Pareto front plot, where each point represents a Pareto optimal solution, the horizontal axis of the point represents the total cost, and the vertical axis of the point represents the total carbon emissions.

[0052] In the NSGA-II algorithm: Decision variables ( ) is mapped to chromosome genes; fitness function for: = * ; This process should follow an elitist retention strategy, that is, retain the non-dominated solutions of each generation to avoid getting trapped in local optima.

[0053] Step S3C: The user selects the Pareto optimal solution.

[0054] In this step, the Pareto front plot is displayed through an interactive visualization interface, and users can select one of the Pareto optimal solutions and the corresponding deployment scheme according to their actual needs.

[0055] The robot deployment schemes corresponding to the Pareto optimal solution set include: Minimum carbon emission option: Sacrificing some economic efficiency to achieve maximum carbon emission reduction; Lowest cost solution: Select the economically optimal solution within carbon emission constraints; Equilibrium solution: Balance the two through a compromise solution.

[0056] Each robot deployment scheme consists of decision variables corresponding to the Pareto optimal solution. The values ​​are composed of .

[0057] Step S4: Based on the selected robot deployment scheme, correct the Pareto optimal solution set.

[0058] Step S4A: Collect actual operating data in the virtual environment in real time.

[0059] A high-fidelity virtual twin is constructed based on physical robots, and the actual operation effect of the selected robot deployment scheme is simulated in the digital twin platform, and the actual operation data in the virtual environment is collected in real time.

[0060] The simulation data includes total carbon emissions data and total cost data.

[0061] The construction of the digital twin platform is based on the physical robot to build a virtual twin, synchronizes the running data (location, energy consumption, fault status) in real time, and simulates the carbon emissions and costs under different deployment schemes to verify the feasibility of the optimization results.

[0062] Step S4A specifically includes the following steps: Based on the selected parameters, a virtual twin identical to the physical robot is constructed, and an environmental map, task queue, and dynamic energy supply model are loaded. Initiate high-fidelity simulation and use the physics engine to calculate the dynamic energy consumption of the robot under typical working conditions, and combine it with real-time grid carbon emission factors to predict the carbon footprint. Output a verification report, including carbon emission time series curves, cost distribution heatmaps, and task completion rate indicators, and indicate the deviation rate between theoretical predictions (actual operating data) and simulated values ​​(simulated data).

[0063] The digital twin platform incorporates an optimization sandbox, which is an isolated testing environment built using containerization technology. Once the Pareto optimal solution set is obtained, multiple candidate solutions can be deployed and run in parallel within the optimization sandbox to simulate long-term performance under identical operating conditions. By comparing the carbon emission time-series curves, cost fluctuations, and failure rate indicators of each solution within the sandbox, the most robust and optimal deployment solution can be selected without interfering with actual production, thereby reducing the technical risks of directly deploying a single solution.

[0064] The theoretical prediction value refers to the result F directly output by the optimization algorithm. Time series curves, heat maps, and completion rate indicators are all verification indicators. They are in-depth decompositions and visualizations of the theoretical prediction value, used to more accurately evaluate the feasibility and risks of the solution.

[0065] Simulated values ​​refer to the results output after the digital twin runs in a virtual environment. They also include the total carbon emissions, total cost, and all indicators such as the carbon emission time series curve, cost distribution heat map, and task completion rate obtained from the simulation.

[0066] Step S4B: Update the parameters of the NSGA-II algorithm online using the DQN algorithm, and then execute step 3 to obtain the corrected Pareto optimal solution set.

[0067] Using carbon emission intensity, energy price, and equipment health as state inputs, and the difference between actual operating data (obtained in step S4) and simulated data (an optimal solution in step S3C) as a reward signal (reward function), the NSGA-II algorithm's parameters such as crossover probability and mutation rate are dynamically adjusted through the Q-learning mechanism to generate a corrected Pareto optimal solution set, forming a "decision-simulation-feedback" closed-loop optimization mechanism.

[0068] Actual operational data is transmitted back in real time through the data acquisition module, triggering a new round of life cycle carbon accounting and dynamic optimization, forming a continuous closed-loop optimization mechanism of "perception-decision-verification-evolution".

[0069] The parameters of the NSGA-II algorithm include: crossover probability and mutation rate.

[0070] In the DQN algorithm: The state space consists of: real-time carbon emission intensity, energy price, and equipment health.

[0071] The scope of action includes: adjusting maintenance strategies and energy allocation ratios.

[0072] Reward function: ; In the formula, and This is a historical benchmark value. , This is the reward coefficient.

[0073] - Total carbon emissions data from actual operation; - Total cost data from actual operating data.

[0074] The DQN algorithm feeds the validation results back into the optimization model: When the verification deviation rate is ≤5% or the user manually confirms, the final deployment plan is output to the physical robot for execution.

[0075] The "verification results" refer to a series of evaluation data and reports output from the digital twin platform after performing high-fidelity simulation of the user's selected initial robot deployment plan. These include carbon emission time-series curves, cost distribution heatmaps, and simulated total carbon emissions.

[0076] "Optimization model" specifically refers to the multi-objective optimization mathematical model constructed in the previous step of this invention, and its standard form is the bi-objective optimization model F defined in the document.

[0077] "Deviation rate" specifically refers to the difference between the baseline value (theoretical value) and the actual value (simulated value) calculated in step S4A.

[0078] The deviation rate (δ) = |(simulated value - theoretical prediction value) / theoretical prediction value| × 100%, which requires both results to be satisfied, i.e., the union of the two results. Regarding the reward function: The simulation data comes from the Pareto optimal solution selected by the user in step S3C. After the user selects a preliminary solution, the system uses these parameters as input to build a virtual twin in the digital twin platform and start a high-fidelity simulation.

[0079] The actual operational data comes from the real data obtained in step S4 after the scheme, which has been verified and corrected by digital twins, is deployed on the physical robot and monitored in real time by IoT sensors and data acquisition modules.

[0080] The reward function does not directly contain the two types of original data, but uses the deviation between the two as the core calculation element.

[0081] Simulation data for the selected scenario is obtained from the digital twin platform to obtain the simulated total carbon emissions. and simulated total cost .

[0082] Obtain actual operational data after deployment from physical robotic systems to obtain the actual total carbon emissions. and actual total cost .

[0083] The specific steps of step S4B are as follows: First, randomly initialize the number of deployments ( ), Path planning Energy type ) and maintenance strategies ( The population solution set is obtained; the offspring population is generated iteratively through crossover and mutation operations, and non-dominated solutions are screened by combining elite retention strategy; finally, the Pareto front plot is output, and the parameter combination and objective function value of each solution are visualized in a two-dimensional coordinate system (carbon emissions - total cost).

[0084] like Figure 5 The principle of digital twin verification and feedback optimization, as shown in step S4, is as follows: A platform for building virtual twins based on physical robots is deployed, which supports a network of physical devices with 5G / TSN. Data is synchronized to the digital twin in real time via the OPC UA protocol. A multi-physics simulation image is generated in the cloud, and edge nodes perform data cleaning and real-time damage monitoring.

[0085] like Figure 5 As shown, the digital twin verification and feedback optimization platform of the present invention constitutes a closed-loop system. Its core modules include: Federated learning aggregation module: By employing a federated learning protocol among digital twins of multiple robots or production nodes, models can be collaboratively trained and optimized without the need to centralize raw data.

[0086] Nash equilibrium-based resource allocator: In multi-robot collaborative scenarios, the allocation of resources (such as charging stations and tasks) is modeled as a non-cooperative game. By solving the Nash equilibrium point, a stable allocation scheme with optimal overall system efficiency and no deviation motivation for individual robots is achieved.

[0087] Uncertainty Quantification Module: Introduces the confidence interval principle to quantify the uncertainty of simulation prediction results, outputs risk boundaries, and provides decision-makers with 'best-worst' scenario analysis.

[0088] Digital Threads: Building a data chain that spans all stages of the entire lifecycle ensures that data from raw material traceability to operation and maintenance can be traced in real time and interconnected, providing a reliable data foundation for full-chain carbon accounting.

[0089] The system employs federated learning to aggregate multi-node models, combining homomorphic encryption and trusted matrix verification to ensure privacy and reliability. It generates Pareto front graphs based on the MOEA / D algorithm, and achieves dynamic resource allocation and conflict resolution through Nash equilibrium and DQN deep reinforcement learning. At the same time, it uses the confidence interval principle to quantify uncertainty and ensure that risks are controllable.

[0090] The system traces the entire lifecycle of the data pool through digital threads, integrates weighted transfer learning to improve cross-scenario adaptability, and finally verifies the results through multi-physics simulation and comparison with real data, with an error of less than 2%. Tests show that this solution can reduce operation and maintenance costs by 15%-20% and improve optimization efficiency by more than 30%. It is suitable for high real-time scenarios such as intelligent manufacturing, realizing closed-loop management from data acquisition and simulation optimization to autonomous decision-making.

[0091] The hardware device of the present invention includes the following modules: Data acquisition module: integrates IoT sensors (energy consumption monitoring, GPS positioning) and external database interfaces to acquire grid carbon emission factors, material database dynamic factors and robot operation data in real time ("actual operation data" in step S4B). Computing core module: Equipped with GPU-accelerated multi-objective optimization algorithms and a digital twin engine; Interactive terminal: Provides a visual decision-making interface and parameter configuration panel, displays the Pareto optimal solution set and carbon emission-cost correlation curve, and supports user preference parameter input; Blockchain Nodes: Deploy private blockchain networks to enable carbon data storage and sharing.

[0092] The operating procedure for this hardware facility is as follows: The data acquisition module acquires carbon data throughout the entire lifecycle, including: carbon emission coefficients during the raw material acquisition stage, energy consumption data during the production and manufacturing process, power consumption during robot operation, fuel usage and environmental parameters (such as temperature, humidity, and task load), mileage information for transportation and delivery, and power consumption for operation and maintenance.

[0093] Simultaneously, it connects to blockchain nodes to obtain carbon emission data from raw material suppliers during resource extraction and manufacturing stages, as well as logistics routes and vehicle energy consumption records during transportation deployment stages.

[0094] Meanwhile, the module also connects to an external dynamic parameter database to obtain market environment parameters such as carbon tax policies, energy prices, and the proportion of renewable energy in real time, providing data support for subsequent optimization calculations.

[0095] An external dynamic factor library updates grid carbon emission factors and material-implied carbon data in real time. After multi-source data cleaning and normalization, a full life-cycle carbon data chain is constructed.

[0096] The core computation module uses the improved NSGA-II algorithm to perform multi-objective optimization calculations based on the collected multi-source data.

[0097] A Pareto optimal solution set is generated through intelligent optimization algorithms. Each solution corresponds to a specific set of process parameters, energy usage plans, and operational strategies, i.e., the number of deployments (…). ), Path planning ), energy type combination ( ) and maintenance strategy function ( This provides users with a diverse range of choices.

[0098] Users can intuitively view the Pareto front through the human-computer interface of the interactive terminal and select the most suitable optimization scheme based on actual needs.

[0099] Once a solution is selected, the system will automatically trigger the digital twin verification module to simulate and verify the solution in a virtual environment, evaluating its feasibility and effectiveness in a real-world operating environment. After successful verification, the optimal solution will be deployed to the actual production system.

[0100] The system establishes a complete data closed-loop feedback mechanism, with actual operational data after deployment being transmitted back to the system database in real time. By comparing and analyzing expected results with actual performance, the system can automatically adjust and optimize model parameters to achieve continuous optimization. This "collection-optimization-verification-deployment-feedback" closed-loop mechanism ensures that the system can dynamically adapt to changes in the external environment, continuously improve optimization results, and provide intelligent decision support for enterprises' low-carbon transformation.

[0101] In one specific embodiment, the above technical process can be represented as follows: A multi-objective intelligent scheduling method that integrates dynamic weight allocation based on information entropy and the NSGA-II optimization algorithm is proposed. By inputting sensor data and system load in real time, the method uses an improved Min-Max3-UC normalization algorithm to process multi-objective indicators and combines the target frequency trend with the uncertainty of information entropy quantification to generate adaptive weights.

[0102] Furthermore, the NSGA-II optimizer is used to perform Pareto solution set search under network constraints, and an executable policy is generated through a deep decision network (DOM).

[0103] Verified by cloud computing scheduling, it can improve task completion rate by 28%, reduce energy consumption by 19%, and support real-time hardware deployment in the loop (error <2%).

[0104] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A robot deployment planning method based on full life-cycle carbon emission assessment and optimization, characterized in that, Includes the following steps: Step S1: Construct a full life-cycle carbon emission total model; Step S2: Construct a multi-objective optimization model: Define decision variables, constraints, and a bi-objective optimization model. ; Step S3: Dynamically adjust the multi-model optimization model The weights are then determined, and the NSGA-II algorithm is used to solve the multi-objective optimization model. Obtain the Pareto optimal solution set for users to select from the Pareto optimal solutions; Step S4: Based on the selected robot deployment scheme, correct the Pareto optimal solution set.

2. The robot deployment planning method based on full life cycle carbon emission assessment and optimization according to claim 1, characterized in that, Step S2 specifically includes the following steps: Step S2A: Define decision variables ; Step S2B: Establish a system based on total carbon emissions throughout the entire life cycle. and based on total cost Bi-objective optimization model : Step S2C: Set constraints.

3. The robot deployment planning method based on full life cycle carbon emission assessment and optimization according to claim 2, characterized in that, Decision variables for: ； in, - Number of deployments; - Energy type Related robot operation paths; - Maintenance strategy functions related to time 𝜏.

4. The robot deployment planning method based on full life cycle carbon emission assessment and optimization according to claim 1, characterized in that, Step S3 specifically includes the following steps: Step S3A: Introduce the entropy weight method to dynamically adjust the weights of economic objectives. Weighting of environmental protection goals ; Step S3B: Solve for the Pareto optimal solution set using the NSGA-II algorithm; Step S3C, user selects Pareto optimal solution.

5. The robot deployment planning method based on full life cycle carbon emission assessment and optimization according to claim 1, characterized in that, Step S4 specifically includes the following steps: Step S4A: Real-time acquisition of actual operating data in the virtual environment; Step S4B: Update the parameters of the NSGA-II algorithm online using the DQN algorithm, and then execute step 3 to obtain the corrected Pareto optimal solution set.

6. The robot deployment planning method based on full life cycle carbon emission assessment and optimization according to claim 5, characterized in that, The reward function of the DQN algorithm: ； in, and This is a historical benchmark value. , This is the reward coefficient; - Total carbon emissions data from actual operation; - Total cost data from actual operating data.

Images