A multi-network adaptive handover optimization algorithm based on vehicle space-time behavior rules

Abstract

This invention discloses a multi-network adaptive switching optimization algorithm based on the spatiotemporal behavior patterns of vehicles. Targeting three types of intelligent vehicles with fixed spatiotemporal operational patterns—unmanned delivery vehicles, intelligent buses, and automated sweeping vehicles—it innovatively constructs a spatiotemporal behavior feature model and a multi-dimensional weighted cost-effectiveness evaluation system, leveraging the heterogeneous multi-network hardware architecture of the vehicles, including multi-cellular modules, WiFi, and satellite. By mining the periodic operational characteristics of vehicles at fixed time periods and locations, and combining this with the stability patterns of the network state at corresponding locations, predictive proactive network switching is achieved. The algorithm integrates four core dimensions—cost, latency, bandwidth, and stability—to calculate the overall cost-effectiveness of the network, dynamically adjusting weights according to different vehicle operational scenarios to differentiate and adapt to various operational needs. This invention significantly improves the stability, economy, and adaptability of communication for intelligent operational vehicles, making it suitable for large-scale commercial deployment of various fixed-track intelligent unmanned vehicles.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicle network communication technology, specifically involving a multi-network adaptive switching optimization algorithm adapted to vehicles with normalized fixed operations such as unmanned delivery vehicles, intelligent buses, and automatic sweeping vehicles. It is applicable to intelligent decision-making and dynamic switching optimization of communication networks in multi-module heterogeneous network fusion scenarios. Background Technology

[0002] Currently, unmanned operational vehicles and intelligent public service vehicles are widely used in urban delivery, public transportation, and municipal cleaning. These vehicles require continuous and stable network communication to support core functions such as data transmission, remote control, autonomous driving perception, and task scheduling. Most mainstream vehicle communication solutions currently employ a single network standard or a simple network optimization switching mode, relying solely on real-time signal strength for network switching decisions, which presents significant technical shortcomings.

[0003] First, a single network cannot simultaneously meet the comprehensive requirements of cost, transmission latency, bandwidth capacity, and operational stability. 4G / 5G cellular networks offer low latency and strong stability but are expensive; WiFi networks are free and low-cost but have fragmented coverage and poor stability; satellite networks provide full coverage but have limited bandwidth and relatively high latency. The advantages and disadvantages of each network are significant, and traditional handover methods cannot achieve optimal adaptation of multiple network resources. Second, existing network handover algorithms rely solely on instantaneous judgments based on real-time network parameters, without considering vehicle operating characteristics. This leads to frequent invalid handovers, handover delays, and network mismatches, which can easily cause vehicle operation interruptions, data packet loss, and remote control failures.

[0004] Unmanned delivery vehicles, buses, and sweepers all exhibit strong spatiotemporal regularity, with highly fixed daily operating routes, working hours, stopping points, and task processes. Furthermore, the network environment (signal strength, bandwidth, latency, and cost) at these fixed spatiotemporal locations shows periodic stability. However, existing technologies do not utilize this core characteristic, making it impossible to achieve predictive and forward-looking network switching. This makes it difficult to balance the overall cost-effectiveness of communication quality and operating costs, resulting in wasted vehicle communication resources and insufficient communication reliability, which hinders the large-scale commercial deployment of intelligent unmanned vehicles. Summary of the Invention

[0005] To address the shortcomings of existing technologies, such as simplistic network switching decisions, unbalanced cost-effectiveness, poor adaptability, failure to utilize vehicle spatiotemporal operational patterns, and frequent ineffective switching, this invention provides a multi-network adaptive switching optimization algorithm based on vehicle spatiotemporal behavior patterns. This algorithm is applicable to three types of vehicles operating in fixed scenarios: unmanned delivery vehicles, intelligent buses, and automated sweeping vehicles. By combining a multi-dimensional weighted cost-effectiveness evaluation system with big data prediction of vehicle spatiotemporal behavior, it achieves intelligent, precise, and proactive switching of heterogeneous networks, balancing communication stability, low latency, high bandwidth, and low cost, significantly improving the overall cost-effectiveness of vehicle communication.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a multi-network adaptive switching optimization algorithm based on vehicle spatiotemporal behavior patterns, deployed on intelligent operating vehicles equipped with multi-network heterogeneous modules, including unmanned delivery vehicles, intelligent buses, and automatic sweeping vehicles. The vehicle hardware is equipped with at least two of the following: multiple cellular network modules, WiFi network cards, and satellite communication modules, enabling parallel online operation of multiple networks, real-time monitoring, and on-demand switching. The algorithm comprises five core modules: vehicle spatiotemporal behavior modeling, real-time network parameter acquisition, multi-dimensional cost-effectiveness weighted evaluation, predictive network switching decision-making, and dynamic weight iterative optimization. The specific steps are as follows: 1. Vehicle Spatiotemporal Behavior Pattern Modeling: Based on historical vehicle operation big data, a vehicle spatiotemporal behavior feature database is constructed to accurately record the vehicle's daily fixed-time operating trajectory, stopping location, and operational status, forming a standardized spatiotemporal mapping model. For unmanned delivery vehicles, daily delivery time, pick-up and delivery points, and dwell time are recorded; for intelligent buses, departure time, stop sequence, and route are recorded; for automated sweepers, cleaning time, area operation range, and fixed-point cleaning duration are recorded. Through big data clustering analysis, the periodic fixed characteristics of vehicle "time-location-operation status" are verified and solidified, while historical network state data of corresponding spatiotemporal points are associated to form a spatiotemporal-network feature association database.

[0007] 2. Real-time Collection of Multi-Network, Multi-Dimensional Parameters: The vehicle terminal collects core performance and cost parameters of all access networks in real time and in parallel. These parameters cover four core indicators: cost, transmission latency, available bandwidth, and operational stability. Cost parameters include unit traffic cost and time-based cost standards; latency parameters include data round-trip latency and transmission jitter; bandwidth parameters include real-time uplink / downlink available bandwidth and peak bandwidth; and stability parameters include signal strength, packet loss rate, and disconnection frequency. All parameters are synchronized in real time to the algorithm decision-making terminal, forming a dynamic network parameter matrix.

[0008] 3. Construction of a Multi-Dimensional Weighted Cost-Effectiveness Evaluation Model: A multi-index weighted cost-effectiveness scoring formula is established, abandoning the single-index selection model. The weights of each index are dynamically configured according to the needs of vehicle operation scenarios, and the real-time cost-effectiveness score of each candidate network is calculated comprehensively. The cost-effectiveness scoring formula is: Comprehensive cost-effectiveness score S = α × S1 + β × S2 + γ × S3 + δ × S4, where S1 is the cost-effectiveness score, S2 is the latency performance score, S3 is the bandwidth capability score, and S4 is the stability score. α, β, γ, and δ are the dynamic weight coefficients of the corresponding indicators, and α + β + γ + δ = 1. The weight coefficients can be adaptively adjusted according to the vehicle operation scenario, prioritizing core operational needs in real time.

[0009] 4. Predictive Network Switching Decision Mechanism: Combining the vehicle's real-time spatiotemporal location, current time, and pre-stored spatiotemporal behavior patterns, the mechanism predicts the vehicle's location changes and operational status over a future preset time period. It matches this with historical network environment characteristics of the corresponding location and anticipates performance fluctuation trends for each network. Based on real-time cost-effectiveness scores and predicted network trends, a tiered switching decision is executed: when the target network's overall cost-effectiveness score exceeds the current network's preset threshold, and the target network's status is predicted to be stable in the future, a smooth network switch is triggered; when the vehicle enters a fixed, low-cost, high-stability area, a low-cost network is prioritized; when the vehicle enters a high-speed, high-signal-fluctuation area, a low-latency, high-stability cellular network is prioritized; and in remote areas without base stations, satellite networks are automatically switched, achieving precise scenario-based adaptation.

[0010] 5. Dynamic weight iterative optimization: The algorithm continuously collects vehicle operation data and network operation data after each network switch, and iteratively optimizes the weight coefficients of each indicator through machine learning. It continuously optimizes the cost-effectiveness evaluation model for different vehicles, different scenarios, and different time periods, adapts to long-term network environment changes and vehicle operation fine-tuning, and ensures the long-term optimal adaptability of the algorithm.

[0011] Furthermore, the multi-network module supports parallel monitoring and seamless switching. During the switching process, a data caching transition mechanism is adopted to avoid data disconnection and packet loss during the switching process, ensuring the continuous transmission of vehicle autonomous driving perception data, scheduling instructions, and monitoring data.

[0012] Furthermore, the initial weight values ​​are configured differently for different vehicle scenarios: unmanned delivery vehicles focus on cost and stability to improve the economics of delivery operations; smart buses focus on latency and stability to ensure real-time scheduling and passenger information transmission; and automatic sweeping vehicles focus on cost and bandwidth to reduce municipal operation and maintenance costs.

[0013] Furthermore, the spatiotemporal-network feature association database supports scheduled updates, automatically reviewing the vehicle's operating trajectory and network status data daily, and matching the update cycle with vehicle operation patterns to ensure continuous accuracy in prediction.

[0014] Compared with the prior art, the present invention has the following significant innovative advantages: 1. Innovative utilization of the spatiotemporal regularity of fixed-operation vehicles: For the first time, a spatiotemporal-network correlation model is constructed for the periodic fixed operation and fixed-location operation characteristics of unmanned delivery vehicles, buses, and sweeping vehicles, so as to realize predictive active network switching, overcome the lag defects of traditional real-time monitoring and passive switching, and reduce network adaptation errors and invalid switching from the root.

[0015] 2. Construct a multi-dimensional weighted cost-effectiveness evaluation system: Break away from the traditional logic of selecting the best single signal, integrate four core dimensions of cost, latency, bandwidth and stability into a weighted score, take into account both communication quality and operating costs, significantly improve the overall cost-effectiveness of vehicle network communication, and effectively reduce the communication cost of long-term operation of unmanned vehicles.

[0016] 3. Adaptable to various heterogeneous network convergence scenarios: Fully compatible with parallel access and intelligent switching of multiple networks such as cellular multi-module, WiFi, and satellite, covering communication needs in all scenarios such as urban roads, business districts and parks, remote road sections, and closed factory areas, solving the shortcomings of single network coverage.

[0017] 4. Scenario-based adaptive optimization capability: Differentiated weights are configured for the operational characteristics of the three types of target vehicles, and dynamic iterative optimization is performed by combining machine learning to adapt to the core needs of different operational scenarios. At the same time, it has strong versatility and scalability and can be adapted to various types of fixed-track intelligent operation vehicles.

[0018] 5. Ensure communication stability and operational continuity: By using a predictive switching and caching transition mechanism, frequent network jitter, disconnection, and data packet loss are avoided, which greatly improves the stability of autonomous driving, remote scheduling, and data uploading of unmanned vehicles and ensures reliable operation of vehicles around the clock. Detailed Implementation

[0019] The present invention will be further described in detail below with reference to specific embodiments.

[0020] Example 1: Adaptation to Unmanned Delivery Vehicle Scenarios Unmanned delivery vehicles operate on fixed routes within parks, communities, and commercial districts during fixed time periods each day, with highly fixed operating trajectories, stop locations, and operating hours. The vehicles are equipped with a dual-network architecture consisting of dual 5G modules and WiFi network cards. The algorithm's initial weight configuration is as follows: cost weight α = 0.4, stability weight δ = 0.3, latency weight β = 0.15, and bandwidth weight γ = 0.15, prioritizing low-cost and highly stable delivery communication needs.

[0021] By modeling spatiotemporal behavior, the system records the delivery locations and routes of trucks daily from 9:00 AM to 6:00 PM, and correlates this with historical network data for each location: stable WiFi coverage and zero cost within the park; low latency and high stability of the 5G network in road driving areas; and congestion and insufficient bandwidth of single 5G modules in densely populated commercial areas. Based on these predictions, the algorithm switches to WiFi in advance when vehicles enter the park for delivery, reducing data costs; switches to the optimal 5G network operator while vehicles are driving on roads, ensuring low latency for driving control; and automatically switches to a backup 5G module in congested commercial areas, ensuring sufficient bandwidth. The entire process uses dynamic optimization through weighted scoring, resulting in a cost reduction of over 35% and a 40% improvement in network stability compared to traditional fixed network modes.

[0022] Example 2: Smart Bus Scene Adaptation The intelligent bus operates according to a fixed timetable and fixed stops, and is equipped with a 4G / 5G dual-mode network and satellite backup network. The algorithm's initial weight configuration is as follows: latency weight β = 0.4, stability weight δ = 0.35, bandwidth weight γ = 0.15, and cost weight α = 0.1, prioritizing the low latency and high stability requirements of real-time scheduling, on-board monitoring, and passenger network.

[0023] Based on spatiotemporal pattern predictions, the algorithm proactively switches to the superior 5G network in urban areas during peak hours when the bus network is congested. In suburban areas, where 4G networks are stable and cheaper, the algorithm automatically switches to 4G. In tunnels and remote areas with weak base station signals, the algorithm automatically switches to satellite networks as a backup. Simultaneously, it dynamically monitors network latency and packet loss rate to ensure uninterrupted transmission of bus dispatch instructions, real-time location data, and video surveillance data, effectively avoiding network lag and instruction delays during peak hours.

[0024] Example 3: Scenario Adaptation for Automatic Sweeping Vehicles The municipal automated sweeper cleans fixed municipal roads and park roads at fixed times every day. Its operating trajectory is fixed, and the operating environment is stable. It is equipped with a single 5G module and WiFi network. Algorithm initialization weight configuration: cost weight α = 0.45, bandwidth weight γ = 0.25, stability weight δ = 0.2, latency weight β = 0.1, prioritizing control of operation and maintenance communication costs.

[0025] The algorithm uses historical data to model and identify areas with full WiFi coverage, such as parks, industrial parks, and municipal compounds, within the cleaning area. When vehicles enter the cleaning area, the algorithm automatically switches to free WiFi networks to upload cleaning data and receive tasks. In areas without WiFi on main urban roads, the algorithm automatically switches to 5G networks to ensure basic operational communication. Through scenario-based network switching, the algorithm minimizes data traffic costs while meeting the communication needs of cleaning operations, significantly improving the cost-effectiveness of municipal operations and maintenance.

Claims

1. A multi-network adaptive switching optimization algorithm based on vehicle spatiotemporal behavior patterns, characterized in that, Deployed on intelligent operation vehicle terminals, the intelligent operation vehicles include unmanned delivery vehicles, intelligent buses, and automatic sweeping vehicles; the vehicles are equipped with multiple heterogeneous communication modules, including at least two of the following: multiple cellular network modules, WiFi network cards, and satellite communication modules, supporting parallel access to multiple networks and real-time parameter monitoring; The algorithm includes the following steps: 1) Construct a vehicle spatiotemporal behavior feature database. Based on historical vehicle operation big data, solidify the periodic operation patterns of vehicles at fixed time periods, fixed locations, and fixed operating states, establish a "time-location-operating state" spatiotemporal mapping model, and associate it with the historical network state data of the corresponding points to form a spatiotemporal-network feature association database. 2) Real-time parallel collection of multi-dimensional parameters of all access networks, including real-time data of four core indicators: cost, transmission latency, available bandwidth, and operational stability, to construct a dynamic network parameter matrix; 3) Construct a multi-dimensional weighted cost-effectiveness evaluation model. Calculate the comprehensive cost-effectiveness score of each candidate network using a weighted scoring formula for cost, latency, bandwidth, and stability. The weights of each indicator can be adaptively configured according to the vehicle's operating scenario. 4) Based on the real-time spatiotemporal location of the vehicle and the pre-stored spatiotemporal behavior patterns, predict the future location changes of the vehicle and the fluctuation trend of the network environment, and perform predictive smooth network switching in combination with the real-time cost-effectiveness score. 5) Based on machine learning, the weight coefficients are dynamically iteratively optimized, and the evaluation model is continuously optimized according to historical switching data and job performance to adapt to dynamic changes in scenarios and network environments.

2. The algorithm according to claim 1, characterized in that, The scoring formula of the multi-dimensional weighted cost-effectiveness evaluation model is: S = α × S1 + β × S2 + γ × S3 + δ × S4; where S is the comprehensive cost-effectiveness score, S1 is the cost-effectiveness score, S2 is the latency performance score, S3 is the bandwidth capability score, and S4 is the stability score; α, β, γ, and δ are the dynamic weight coefficients of the corresponding indicators, and α + β + γ + δ = 1.

3. The algorithm according to claim 1, characterized in that, Different initial weights are configured for different vehicle scenarios: unmanned delivery vehicles are configured with high weights for cost and stability; intelligent buses are configured with high weights for latency and stability; and automatic sweeping vehicles are configured with high weights for cost and bandwidth.

4. The algorithm according to claim 1, characterized in that, The predictive network switching mechanism specifically includes: predicting the network status of the target area in advance based on the vehicle's preset operating trajectory and spatiotemporal patterns, and completing the network pre-switching before the vehicle enters the target area to avoid the lag of passive switching; the switching process adopts a data caching transition mechanism to achieve seamless switching and prevent data packet loss and transmission interruption.

5. The algorithm according to claim 1, characterized in that, The spatiotemporal-network feature association database supports daily automatic updates, reviews the vehicle operation trajectory and network operation data of the day, and iteratively optimizes the association accuracy between spatiotemporal patterns and network status.

6. The algorithm according to claim 1, characterized in that, The heterogeneous network switching logic includes: prioritizing the switching to low-cost WiFi networks in closed parks and fixed parking locations; prioritizing the switching to low-latency, high-stability cellular networks in high-speed driving and real-time scheduling scenarios; and automatically switching to satellite networks as a backup in remote areas without base stations, achieving optimal network adaptation across all scenarios.