The research team set out to solve a problem that has long frustrated engineers working on off-road hybrid vehicles. In mountainous orchards, a machine’s power demand is nonstationary, meaning it does not follow stable statistical patterns and can change abruptly as the vehicle crests a ridge or turns onto a steeper grade. Traditional rule-based energy management strategies, which dispatch power according to fixed thresholds and simple logic, offer essentially no preview of what the terrain ahead will demand. Instantaneous optimization strategies fare little better, because they minimize fuel use only for the present instant without regard to what comes next. Model predictive control, or MPC, does look ahead by solving an optimization problem over a rolling time horizon, and it can feed prediction errors directly into actuator commands. But that capability comes at a price: MPC imposes substantial online computation, which is a serious burden for the resource-constrained embedded controllers typical of agricultural machinery.

The researchers’ solution begins with a forward model of a power-split hybrid agricultural machine. The architecture comprises a combustion engine, two electric machines, a traction battery, and a planetary gear set, the latter allowing power from the engine and the electric machines to be split and recombined mechanically. This kind of power-split layout is attractive because it lets the engine operate near its sweet spot of efficiency while the electric machines absorb or supply the difference between what the engine produces and what the wheels demand. The team built a mathematical description of how power flows through this system, including two selectable transmission ratios and the physical constraints on every actuator, so that any control strategy could be tested against a faithful representation of the machine.

At the heart of the new strategy is a concept the authors call a bounded load-risk indicator. The machine looks a short distance into the future, gathering predictions of upcoming vehicle speed and road gradient. These predictions are then normalized, weighted, and saturated, a mathematical process that caps the indicator at a fixed upper bound so that no single alarming prediction, or any burst of prediction noise, can dominate the control decision. The indicator essentially condenses the near future into a single number that expresses how demanding the road ahead is likely to be. Crucially, rather than feeding this number directly into low-level torque calculations, where a prediction error would propagate straight into the actuators, the designers use it to modulate higher-level decision boundaries.

This soft decoupling of prediction and execution is the strategy’s defining trick. The load-risk indicator modifies three things: the dynamic safety boundary of the battery’s state of charge, the selection of the power-source mode, and the switching threshold between the two transmission ratios. In other words, if the indicator senses a demanding climb ahead, it nudges the battery’s usable range, shifts the mode selection toward configurations better suited to heavy load, and adjusts when the transmission jumps between ratios. The actual torque allocation is then carried out by a computationally efficient online scheme that coordinates engine high-efficiency operation, the two transmission ratios, and actuator constraints. Because prediction errors only reshape decision boundaries instead of commanding torques directly, the machine retains the benefit of preview while limiting how far a bad prediction can propagate into the hardware.

To judge how well this approach performs, the team needed a gold standard. They turned to dynamic programming, or DP, an offline optimization technique that, given a complete driving cycle in advance, computes the theoretically best possible power split at every moment. DP cannot run in real time on a machine because it requires perfect foreknowledge of the entire route, but it establishes an unbeatable benchmark. Over three test cycles, labeled C1 through C3, the new predictive-feedforward strategy achieved between 93.29 and 95.59 percent of the DP performance. That is a striking result for a real-time method that knows nothing about the route beyond a short prediction horizon.

The comparison against other real-time strategies is equally telling. The new approach exceeded MPC by 1.41 to 3.19 percentage points in fuel-related performance, and it beat adaptive equivalent-consumption minimization, a well-known real-time strategy that adapts a fuel-equivalence factor for battery use, by 3.62 to 4.99 percentage points. Meanwhile, the computational advantage was dramatic. The mean execution time of the new strategy was between 0.128 and 0.191 milliseconds per control step, which is 93.54 percent lower than that of MPC. For an embedded controller with limited processing resources, that difference is the gap between a strategy that fits comfortably in real time and one that strains the hardware.

Battery health is another central concern for any hybrid machine, because deep or erratic battery cycling shortens component life and raises operating costs. In the tests, the terminal state of charge remained near 0.600, indicating that the machine finished its work with the battery in a healthy, sustainable condition rather than having drained it to compensate for poor engine coordination. More importantly, the strategy proved robust under stress. When the researchers injected prediction noise and abrupt disturbances in the equivalent rolling resistance, simulating the sudden changes a machine might encounter on uneven orchard lanes, no constraint was violated. The state of charge stayed within bounds, battery current stayed within limits, and the torque of the power sources never exceeded what the actuators could deliver.

The practical implications extend beyond the test bench. Hybrid agricultural machines operating in mountain orchards typically run at low speed under high load, a regime in which the engine’s efficiency map is unforgiving and every percentage point of fuel savings matters for both cost and emissions. A strategy that runs in a fraction of a millisecond per step can be deployed on the modest embedded controllers that agricultural manufacturers actually fit to their machines, without requiring automotive-grade computing hardware. The bounded feedforward design also offers a template for other off-road hybrid applications, from forestry machines to vineyard tractors, where terrain varies continuously and full MPC is too expensive to compute online.

The authors are careful to delineate the limits of their findings. The conclusions apply to maintained, firm orchard lanes with low track slip, the kind of well-kept paths found in managed orchards. Soft-soil operation, which involves high traction demands, severe sinkage of the wheels or tracks, or substantial slip, requires further validation before the strategy can be recommended for those conditions. Within its validated domain, however, the study demonstrates that a machine can enjoy the foresight of prediction without paying the computational price of predictive control. By saturating its view of the future and letting that view reshape decision boundaries rather than command torques, the predictive-feedforward approach retained short-horizon preview while limiting direct error propagation, providing a practical real-time path for hybrid agricultural machines working the hills.

Subject of Research: A predictive-feedforward energy management strategy for power-split hybrid agricultural machines in mountainous orchards

Article Title: Predictive-feedforward energy management for a hybrid agricultural machine operating in mountainous orchards

Article References: Li, Z., Huang, H., Dong, L., Zhang, X., Zhou, R., Li, B., & Yu, B. (2026). Predictive-feedforward energy management for a hybrid agricultural machine operating in mountainous orchards. PLOS One, 21(10), e0360955. https://doi.org/10.1371/journal.pone.0360955

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0360955

Keywords: hybrid vehicles, agricultural machinery, energy management, predictive control, mountain orchards, battery state of charge, Predictive-feedforward, energy, management, hybrid, agricultural, machine

The research, led by Ahmed Mashaly of Mansoura National University, together with Ragab A. El-Sehiemy of Kafrelsheikh University and Mohamed Elgohary of Delta University for Science and Technology, addresses a persistent gap in the smart agriculture literature. According to the authors, most existing studies concentrate either on IoT monitoring platforms or on AI prediction models, but rarely integrate both into a single, coherent system. Their framework does exactly that, coupling a long short-term memory (LSTM) neural network for predicting irrigation water requirements with a complete renewable energy management layer that includes photovoltaic generation, battery state-of-charge (SOC) management, and a maximum power point tracking (MPPT) algorithm.

At the heart of the predictive engine sits the LSTM network, a specialized form of recurrent neural network designed to capture long-range temporal dependencies in sequential data. That architecture is well suited to agriculture, where irrigation demand depends not just on today’s weather but on patterns of solar irradiance, temperature, and humidity unfolding over days. Trained and tested on real environmental data drawn from NASA’s POWER database for the New Delta region, the LSTM classified the operating states of the irrigation pump with an accuracy of 96.11 percent. Its recall of 0.9820 indicates that it rarely misses a genuine irrigation event, while its precision of 0.9100 shows that false alarms are comparatively uncommon. The resulting F1-score of 0.9446 balances those two qualities into a single figure of merit.

Those numbers matter because the model was not operating in isolation. The team benchmarked the LSTM against two widely used alternatives, an artificial neural network (ANN) and a support vector machine (SVM), and found the LSTM outperformed both. Robustness testing across multiple independent training runs showed stable performance, an important consideration for any system intended to run autonomously in the field, where retraining conditions vary and overfitting to a single dataset can quietly degrade reliability.

The energy side of the framework is equally deliberate. A photovoltaic system, whose output power depends on solar irradiance and panel efficiency, feeds the irrigation pump through an MPPT algorithm that continuously adjusts the operating point to extract the maximum available power from the panels. A battery storage subsystem, tracked through its state of charge, buffers the mismatch between when the sun shines and when the crops need water. Rather than running pumps at fixed speed, the framework adapts pump speed to actual demand, which is where the dramatic 74.74 percent reduction in motor energy consumption comes from. In conventional irrigation, motors often run at full power regardless of need; adaptive control trims that waste.

Water savings arise from a different mechanism. Instead of irrigating on a rigid timetable, the system predicts how much water the crops will actually consume and schedules irrigation accordingly, using a soil-moisture error term to correct deviations between predicted and measured conditions. Against an assumed conventional fixed-schedule baseline, the simulated annual water saving reached 28.57 percent. For a country like Egypt, which faces severe water scarcity and is developing the New Delta as new agricultural land, savings of that magnitude could translate into substantially more farmland served by the same water allocation.

The researchers also stress-tested their design. A sensitivity analysis examined how the system responds to moderate variations in environmental and operating conditions, and the model remained stable. That resilience is critical for real-world deployment, where sensor noise, weather variability, and component aging are the norm rather than the exception. The framework further includes a conceptual IoT communication architecture, envisioning field sensors, controllers, and remote monitoring linked through technologies such as GSM and long-range wide area networks, so that the intelligence demonstrated in simulation could eventually govern physical hardware.

It is worth emphasizing what the study does and does not claim. This is a simulation framework, validated against real environmental data but not yet demonstrated on deployed hardware in a working field. The water and energy savings are estimates relative to a modeled baseline, and the authors position the work explicitly as a promising start toward future hardware implementations of intelligent, solar-powered irrigation devices. Even so, the integration is what sets the work apart: prediction, power management, storage control, and communication are designed together rather than bolted together after the fact.

The broader implications extend beyond Egypt. Agriculture consumes the majority of the world’s freshwater withdrawals, and solar-powered irrigation is expanding rapidly across regions where grid electricity is unreliable or absent. A framework that pairs machine learning forecasts with renewable energy management could help farmers squeeze more output from both scarce inputs, water and sunlight. The MATLAB environment also makes the approach reproducible and adaptable, allowing other researchers to plug in local climate data and re-tune the models for their own conditions.

From the sun-scorched fields of the New Delta to the growing literature on AI-driven smart farming, the study offers a template for how simulation can de-risk the next generation of agricultural technology. If future hardware trials confirm the simulated gains, the combination of LSTM forecasting, adaptive pump control, and solar energy management could become a cornerstone of sustainable irrigation in water-stressed regions worldwide.

Subject of Research: An AI-IoT simulation framework using LSTM networks for smart solar-powered irrigation and renewable energy management

Article Title: A MATLAB-based AI-IoT simulation framework for LSTM-driven smart solar irrigation and renewable energy management

Article References: Mashaly, A., El-Sehiemy, R. A., & Elgohary, M. (2026). A MATLAB-based AI-IoT simulation framework for LSTM-driven smart solar irrigation and renewable energy management. Scientific Reports, 16(1), Article 31855. https://doi.org/10.1038/s41598-026-74935-0

Image Credits: AI Generated

DOI: 10.1038/s41598-026-74935-0

Keywords: smart irrigation, LSTM, artificial intelligence, Internet of Things, photovoltaic systems, renewable energy, MATLAB, water efficiency, maximum power point tracking, battery state of charge, smart agriculture, Egypt

The core of the system is a two-layer architecture that separates the geometry of pathfinding from the judgment of how cautious to be. The lower layer generates candidate trajectories using the A* algorithm, a classic graph-search method dating back to a landmark 1968 paper by Hart, Nilsson and Raphael, which finds minimum-cost paths by exploring the most promising branches of a grid first. What makes the Inha team’s implementation distinctive is the structure of its cost function. Every candidate path is scored by a weighted combination of two terms: a distance term that favors shorter routes and shorter required times of arrival, and a risk term that penalizes exposure to hazards. Those hazards come in two flavors, static and dynamic. Static risk is the urban landscape itself, the buildings that a low-altitude air taxi must thread between. Dynamic risk is the weather, specifically wind and gust severity, which can change from one minute to the next and turn a safe corridor into a dangerous one.

The upper layer, and the true innovation of the study, is the AI pilot itself: a reinforcement-learning agent trained with the deep deterministic policy gradient algorithm, or DDPG, a technique introduced by Lillicrap and colleagues in 2016 that excels at continuous control problems. The agent’s single job is to choose a continuous weighting factor, denoted lambda, which controls the relative influence of the distance and risk terms inside the A* cost function. In other words, the AI does not draw the path directly. Instead, it decides how much the path planner should care about speed versus safety, given the current operational context. When the battery is healthy and the air is calm, the agent can push lambda toward efficiency, letting the vehicle fly direct, fast routes. When the state of charge falls or gusts intensify, the agent shifts the weighting toward risk avoidance, and the planner responds by carving out longer but safer detours around buildings and turbulent zones.

One design decision stands out as particularly consequential. Battery state-of-charge and wind information are fed to the policy as continuous observations, not as discrete emergency flags. This choice matters because it changes the character of the system’s behavior at the boundary between normal and abnormal flight. A flag-based system, in which an emergency is declared when some threshold is crossed, tends to produce abrupt, discontinuous responses: the trajectory is one thing before the threshold and something quite different after it. The learned policy, by contrast, produces a lambda that responds smoothly and monotonically to gradually degrading conditions. As the battery drains cell by cell or the wind stiffens knot by knot, the safety weighting tightens proportionally, and the planned path bends incrementally away from hazard. Yet the researchers report that the same policy still reacts decisively when a genuine emergency arrives without warning, snapping to a strongly safety-dominant configuration when the situation demands it.

Validation came through two complementary evaluation strategies. The first was a Monte Carlo benchmark, in which the AI-assisted system was pitted against rule-based baselines and against a planner with a fixed lambda value across many randomized scenarios. Monte Carlo testing is the standard way to expose brittleness in autonomous systems, because it samples the space of possible conditions rather than testing a handful of hand-picked cases. The second was a controlled context-sweep case study, in which a single operational variable, such as battery level or wind severity, was systematically varied while the system’s responses were recorded. This sweep allowed the team to verify the smooth, monotonic behavior of the weighting factor directly, confirming that the agent’s decisions track the operational context in an interpretable way rather than jumping erratically between modes.

Performance was judged on three metrics that map cleanly onto what a passenger would actually feel and what an operator would actually care about. Path length measures efficiency, the extra distance flown to reach the destination. Risk exposure quantifies how much of the flight is spent near buildings or in severe wind and gust conditions. Trajectory smoothness captures how gentle the resulting path is, a property with direct consequences for passenger comfort, vehicle dynamics and control authority. Smoothness is not a cosmetic concern; certification frameworks for piloted aircraft, including the long-standing military flying-qualities specification MIL-F-8785C, treat handling qualities as central to airworthiness, and a replanning system that produced jagged, erratic paths would be unusable in practice regardless of how safe it was on paper.

The study situates itself within a decade of growing interest in flight-deck decision support for advanced air mobility. NASA researchers, including Karr, Ballin, Barrows and colleagues, have developed autonomous operations planners and flight path management automation specifically for high-density urban environments, and flight evaluations of such systems have already been conducted. The NASA urban air mobility maturity level scale, described by Goodrich and Theodore in 2021, provides the community’s roadmap for how these operations might scale from today’s limited demonstrations to high-density fleets. What the Inha work adds to this lineage is the learned, context-sensitive middle layer: rather than encoding the safety-efficiency trade-off as a fixed engineering choice made in advance, the system lets a trained policy make that choice in flight, informed by the actual state of the vehicle and the atmosphere.

The choice of DDPG as the learning algorithm is also worth appreciating on its own terms. Reinforcement learning agents come in many forms, but many of the most successful ones, including the policy-gradient methods behind celebrated game-playing systems, operate over discrete action spaces. Trajectory weighting does not fit that mold. The appropriate balance between speed and safety is not one of several options but a dial with infinitely many settings, and DDPG was designed precisely for such continuous action spaces, using an actor network to propose actions and a critic network to evaluate them. Related work has already shown the algorithm’s usefulness in safety-critical planning, including an improved DDPG model for emergency fire-escape path planning published in 2024, and the Inha team’s application extends that idea into the three-dimensional, hazard-rich environment of urban airspace.

The implications for the emerging air taxi industry are considerable. Certification authorities and operators alike have struggled with the question of how much autonomy to trust in the cockpit, and a system like this one offers a middle path: the AI does not fly the vehicle, it advises the trajectory planner, adjusting a single, physically meaningful parameter that a human pilot or ground operator could inspect and, in principle, override. The researchers describe their AI pilot as a decision-support tool for both gradual context changes and emergency responses during urban air mobility operations, a framing that keeps the human in the loop while offloading the relentless, second-by-second arithmetic of risk and distance. The work was supported by the National Research Foundation of Korea, and the authors report no competing interests.

Challenges remain before such a system carries passengers. The study validates the concept in simulation, and the gap between Monte Carlo benchmarks and flight test in a real city, with real air traffic, real sensor noise and real regulatory scrutiny, is the widest gap in aviation. The authors note that the data supporting the findings are available upon reasonable request, which will help other groups scrutinize and extend the approach. Still, the conceptual contribution is clean and potentially durable: treat the safety-efficiency trade-off not as a design constant but as a learned function of the operational context, and let a continuous, smoothly responding policy turn that function into concrete flight paths. As fleets of electric vertical-takeoff vehicles move closer to routine operation over the world’s cities, the flights that matter most will not be the ones that go according to plan. They will be the ones that do not, and the machines that can quietly, smoothly find a better way home may prove to be the technology that finally makes urban flight trustworthy.

Subject of Research: AI-based emergency trajectory replanning for urban air mobility vehicles

Article Title: AI-Based Pilot Assistance System for Emergency Trajectory Replanning in Urban Air Mobility

Article References: Yu, J., & Choi, K. (2026). AI-Based Pilot Assistance System for Emergency Trajectory Replanning in Urban Air Mobility. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01258-9

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01258-9

Keywords: urban air mobility, AI pilot, reinforcement learning, DDPG, A* algorithm, trajectory replanning, flight path management, safety-efficiency trade-off, wind hazards, battery state of charge, path planning, decision support