Electric Vehicle Routing Optimized for Dynamic Weather and Traffic Conditions
In the rapidly evolving landscape of urban logistics, electric vehicles (EVs) are increasingly seen as a cornerstone of sustainable freight transportation. As cities worldwide push toward decarbonization, last-mile delivery operations face mounting pressure to reduce emissions while maintaining efficiency and reliability. However, despite the growing adoption of electric fleets, many operational models still rely on oversimplified assumptions—such as constant vehicle speed and fixed energy consumption—that fail to reflect real-world complexities. A groundbreaking study conducted by researchers at Jimei University in Xiamen, China, has now introduced a more realistic and cost-effective approach to electric vehicle routing by integrating dynamic environmental variables into route optimization.
Led by postgraduate researcher Zhou Ni, under the supervision of Associate Professor Wang Wen from the College of Navigation at Jimei University, the team developed an advanced routing model that accounts for fluctuating driving speeds caused by weather conditions and traffic patterns throughout the day. Published in the Journal of Jimei University (Natural Science Edition), their work presents a comprehensive framework for optimizing delivery routes for battery-swapping electric vehicles under time-varying conditions, offering practical insights for logistics companies aiming to enhance both economic performance and service quality.
The research addresses a critical gap in current EV routing literature. While numerous studies have explored electric vehicle path planning, most assume uniform travel speeds and consistent energy usage per kilometer. These static models may yield theoretically optimal solutions but often fall short when applied in real-world scenarios where rain, snow, fog, or rush-hour congestion significantly impact vehicle performance. By contrast, the new model explicitly incorporates the influence of external factors such as weather and peak traffic periods on driving speed, which in turn affects energy consumption and overall delivery costs.
At the heart of the innovation is the recognition that vehicle speed is not a fixed parameter but a variable shaped by environmental dynamics. Drawing from prior research on how adverse weather reduces average travel speeds, the team quantified these effects using empirical data. For instance, heavy rainfall can reduce urban driving speeds by up to 30%, while snow and dense fog lead to even greater slowdowns due to reduced visibility and road slipperiness. These variations are modeled through a speed adjustment factor (θ), representing the percentage reduction in average speed based on prevailing conditions.
Crucially, the model links this variable speed directly to energy consumption. Rather than assuming linear energy use, it adopts a quadratic function derived from empirical studies showing that energy expenditure per kilometer increases with speed—but only up to a point. Beyond approximately 50 km/h, aerodynamic drag begins to dominate, causing power demand to rise sharply. This non-linear relationship means that small changes in speed can have disproportionate impacts on battery drain, making accurate speed modeling essential for predicting range and determining optimal charging—or in this case, battery swapping—strategies.
The study focuses specifically on battery-swapping electric logistics vehicles, a technology gaining traction in China and other parts of Asia. Unlike conventional plug-in charging, which can take tens of minutes to hours, battery swapping allows drivers to exchange depleted batteries for fully charged ones in just a few minutes, minimizing downtime. This capability makes it particularly suitable for high-frequency urban delivery operations where time is money.
However, efficient use of swap stations requires intelligent planning. Simply placing swaps along a route without considering timing, load, and energy levels could result in unnecessary detours or missed deliveries. The Jimei University team’s model integrates all these elements: vehicle load capacity, customer time windows, battery state of charge, and the locations of available swap stations. It ensures that each vehicle departs the depot fully charged, services a sequence of customers within their requested time frames, performs swaps when needed, and returns to base—all while minimizing total operational cost.
That cost includes three key components: fixed vehicle usage fees, electricity expenses, and time-based penalties. The latter reflects the financial consequences of early or late arrivals. If a driver arrives before a customer’s preferred window, they must wait, incurring idle time costs. Arriving late triggers penalty fees, potentially damaging client relationships. The model treats both waiting and tardiness as undesirable outcomes, assigning them monetary values to be minimized in the optimization process.
To solve this complex multi-constraint problem, the researchers employed a genetic algorithm—a heuristic search method inspired by biological evolution. Known for its ability to explore large solution spaces efficiently, the genetic algorithm starts with a population of randomly generated routes and iteratively improves them through selection, crossover, and mutation operations. High-performing routes (those with lower total costs) are more likely to pass their “genes” (route segments) to the next generation, gradually converging toward near-optimal solutions.
The implementation used real-number encoding rather than binary strings, allowing for smoother representation of route sequences. Selection was performed via roulette wheel sampling, ensuring diversity in the gene pool while favoring fitter individuals. Single-point crossover and random mutation were applied at low rates to balance exploration and exploitation, preventing premature convergence to suboptimal results.
Testing was conducted using a simulated urban delivery network comprising one central depot, 30 customer nodes, and multiple battery swap stations. Real geographic coordinates were used to calculate inter-node distances, and customer demands and service time windows were assigned based on typical parcel delivery profiles. The simulation ran under four distinct weather scenarios: sunny, rainy, snowy, and foggy conditions, each affecting vehicle speeds differently.
Results demonstrated clear advantages of the dynamic model over traditional fixed-speed approaches. In sunny conditions, with minimal traffic interference, the optimized routes achieved a total cost of ¥1,865. Under rain, the cost rose to ¥2,201.60 due to slower speeds and higher energy consumption; snow increased it further to ¥2,022.30, while fog—the most disruptive condition—pushed the total to ¥2,426.80. These figures underscore the sensitivity of logistics economics to environmental variability.
More importantly, the structure of the optimal routes changed significantly across conditions. In good weather, vehicles could afford slightly longer paths if they allowed better clustering of nearby customers or avoided congested zones later in the day. In poor weather, shorter, more direct routes became preferable to minimize exposure to delays and excessive battery drain. Some customers who received early-morning deliveries in sunny conditions were rescheduled to midday slots during rain, reflecting adjusted arrival predictions based on reduced cruising speeds.
A comparative analysis revealed another key insight: the so-called “most energy-efficient” speed of around 54 km/h—where rolling resistance and air drag balance to minimize kWh/km—is not always the best choice in practice. When average speeds were constrained below 50 km/h (typical in city environments), the dynamic model outperformed any fixed-speed strategy. Only when operating above 50 km/h did maintaining a steady pace near the theoretical efficiency peak yield lower costs. Given that urban delivery vehicles rarely sustain such speeds over extended periods, the adaptive, condition-responsive approach proved superior in nearly all tested cases.
One of the most compelling findings was the degree to which time penalties influenced routing decisions. Because late deliveries incurred a cost of ¥0.50 per minute, the algorithm prioritized punctuality even at the expense of marginally higher energy use. This trade-off highlights the importance of integrating customer satisfaction metrics into logistical planning. A delivery might technically succeed—goods arrive, battery remains sufficient—but if it disrupts the recipient’s schedule, the broader value proposition suffers.
Moreover, the model accounted for the fact that arriving too early also carries hidden costs. Although no explicit fee was set for waiting (λ₁ = 0), prolonged idling wastes labor hours and ties up capital equipment. The system therefore favored routes where arrival times aligned closely with the start of service windows, maximizing asset utilization without risking penalties.
From a managerial perspective, the implications are significant. Logistics firms deploying electric fleets can no longer treat routing software as a one-size-fits-all tool. Instead, they must adopt systems capable of real-time adaptation to changing conditions. Integrating live weather feeds, traffic updates, and predictive analytics would allow dynamic rerouting en route, further enhancing resilience and cost control.
The study also reinforces the strategic value of well-placed battery swap infrastructure. With swap times averaging just a few minutes, these stations function as force multipliers for EV fleets. But their effectiveness depends on accessibility and integration into route planning. The model shows that even with abundant swap options, inefficient placement relative to delivery clusters can negate the benefits. Future investments should therefore align station deployment with high-density delivery corridors and anticipated demand patterns.
While the current model assumes unlimited battery availability at swap points and no queuing delays—a reasonable approximation given high service levels in modern facilities—future iterations could incorporate congestion effects. During peak delivery hours, popular swap stations might experience temporary bottlenecks, adding unpredictability. Modeling such stochastic elements would make the system even more robust.
Another area for expansion involves multi-depot networks. The present study focuses on a single distribution center, common in smaller cities or regional hubs. Larger metropolitan areas often operate decentralized fulfillment centers to shorten delivery radii. Extending the model to coordinate multiple depots, possibly with shared vehicle pools, would reflect the complexity of national logistics operators.
Additionally, incorporating reverse logistics—handling returns, exchanges, or recyclable packaging—could provide a fuller picture of urban freight flows. Many e-commerce deliveries now include return labels, turning one-way trips into round-trip obligations. Planning for bidirectional cargo movement adds another layer of complexity but offers opportunities for consolidation and improved resource use.
Despite these potential enhancements, the existing framework already delivers actionable intelligence. Its validation through repeated simulations across diverse weather types confirms its stability and adaptability. The genetic algorithm consistently converged within 200 generations, producing reliable solutions in acceptable computation time on standard hardware. This efficiency suggests scalability to larger datasets, including hundreds of customers and dozens of vehicles, with only modest increases in processing requirements.
For industry practitioners, the takeaway is clear: embracing dynamic variables isn’t merely an academic exercise—it translates directly into bottom-line savings and enhanced service reliability. Companies that continue relying on outdated, static routing logic risk inefficiencies that accumulate over thousands of daily trips. Those adopting smarter, responsive models stand to gain competitive advantage through reduced fuel (electricity) bills, fewer penalty charges, and higher customer retention.
Policy makers, too, can draw lessons from this research. Urban planning decisions—such as road design, traffic signal coordination, and public transit priority lanes—affect commercial vehicle speeds and, by extension, energy use and emissions. Data-driven models like the one developed at Jimei University can inform infrastructure investments by quantifying the downstream impacts of mobility policies on freight efficiency.
Furthermore, incentives for clean delivery technologies should consider operational realities. Subsidies for EV purchases are important, but equally vital is support for enabling infrastructure—like standardized, interoperable battery swap networks—and digital tools that maximize vehicle productivity. Without smart routing, even the most advanced electric vans may underperform compared to optimized internal combustion engine fleets.
Education and training programs for logistics professionals should also evolve. Understanding the interplay between speed, energy, and scheduling will become increasingly crucial for dispatchers, fleet managers, and supply chain analysts. Academic institutions like Jimei University play a vital role in bridging theory and practice, equipping future leaders with the analytical skills needed to navigate the electrified logistics era.
Looking ahead, the integration of artificial intelligence and machine learning could elevate this work further. Predictive models trained on historical delivery data, combined with real-time sensor inputs, could anticipate disruptions before they occur. Autonomous decision-making systems might adjust routes on the fly, responding to sudden storms or unexpected road closures faster than human operators.
Yet, even without full automation, the principles established in this study offer immediate benefits. By acknowledging that the real world is inherently variable, and designing systems accordingly, organizations can build more resilient, efficient, and sustainable delivery networks.
As global supply chains undergo transformation amid climate imperatives and technological advances, innovations like this represent more than incremental improvements—they signify a shift in mindset. Efficiency is no longer measured solely by distance traveled or packages delivered per hour, but by the holistic balance of cost, timeliness, environmental impact, and customer experience.
Zhou Ni, Wang Wen, Xue Han, and Chen Qiong’s contribution exemplifies how rigorous scientific inquiry can address pressing industrial challenges. Their model does not merely optimize a route—it redefines what optimization means in the context of modern urban logistics.
Zhou Ni, Wang Wen, Xue Han, Chen Qiong, College of Navigation, Jimei University. Published in Journal of Jimei University (Natural Science Edition), DOI: 10.19715/j.jmuzr.2024.01.06