Unmanned Aerial Vehicles (UAVs) are playing an increasingly vital role in applications, such as environmental monitoring, disaster relief, and logistics. Path planning is crucial for these missions; however, generating multi-objective trajectories that satisfy dynamic constraints while optimizing path length, smoothness, and energy consumption remains a challenge. Although Deep Reinforcement Learning (DRL) offers an end-to-end solution, it often suffers from low sample efficiency and blind exploration. To address these issues, we propose a novel UAV Three-Dimensional (3D) path planning method based on Artificial Potential Field (APF) guidance and an LSTM-enhanced Soft Actor–Critic (SAC) algorithm (APF–LSTM–SAC). First, to overcome the inefficiency of random exploration, we utilize APF as heuristic guidance, whose state-dependent forces are integrated with the SAC policy to accelerate convergence. Furthermore, we introduce an LSTM structure into the Actor–Critic network which enables the agent to generate smoother and more dynamically feasible flight trajectories by capturing the temporal dependencies in the state sequence. Comprehensive tests in complex 3D simulation environments demonstrate that the proposed method outperforms pure DRL algorithms. Specifically, it achieves improvements in convergence speed and energy consumption of at least 15.14
This study presents a surrogate-based global optimization to determine the optimal Individual Blade pitch control (IBC) schedule with multiple-harmonic inputs to simultaneously reduce rotor vibration and improve aerodynamic performance of a generic lift-offset coaxial rotor at 250 knots. An aeromechanics model of the lift-offset coaxial rotor is developed and analyzed using CAMRAD II, while surrogate modeling and optimization are performed with MATLAB. A Kriging surrogate model is built based on the CAMRAD II analysis results for parametric study using the selected IBC actuation amplitudes (AN) and control phase angles (φN) of 2P and 3P IBC input combinations (θIBC=2P/A2/φ2 + 3P/A3/φ3). A weighted Tchebycheff scalarization approach, which incorporates varying weight factors between rotor vibration (VI) and performance (L/De), is integrated with a genetic algorithm to perform global optimization. A death-penalty strategy is applied to retain only feasible candidates satisfying both vibration reduction and aerodynamic performance improvement criteria. Among the optimal cases with the weighted Tchebycheff scalarization approach, the minimum vibration solution achieves a maximum VI reduction of 57.49
This study proposes a mission optimization framework for the datalink-enabled anti-ship cruise missile (ASCM). Mission control variables include target assignment, impact course, and simultaneous time-on-target (STOT) offsets. Mission performance is evaluated using a probabilistic layered defense survivability model with Monte Carlo–based estimation. We solve the resulting time-bounded mission replanning problem using particle swarm optimization (PSO) with a fast–fine fitness evaluation strategy and a repair-based procedure for operational constraints, including inter-missile collision avoidance. Simulation results show improved mission effectiveness and survivability compared with a densest-STOT baseline. This framework provides a methodological basis for practical dynamic mission control of ASCMs in network-centric warfare environments.
This study develops a problem formulation for Weapon–Target Assignment (WTA) optimization against missiles with chemical and biological warhead. We incorporate not only the predicted damage caused by the survived target, but also that of the debris from the intercepted target, which has not been taken into account in existing WTA formulations, into the objective function. To relax the nonlinearity, we introduce a new binary variable and show that the problem with the binary variable is an integer linear programming problem. To effectively solve, this paper applies an exact algorithm, that is called Marginal Return-Based Column Enumeration (MRCE) algorithm. Numerical simulations comparing MRCE with a Genetic Algorithm (GA) demonstrate that MRCE achieves higher solution quality with significantly reduced computational time. Additional simulations replicating surface-to-surface engagement scenarios validate the algorithm’s effectiveness in minimizing damage and optimizing missile intercepts. The simulation results confirm that the proposed problem formulation and exact approach provides significantly different, but more reasonable quality of solution compared with that of the typical WTA problem formulation.
Effective protection of high-value assets against electro-optical guided missiles requires precise spatiotemporal coordination of smoke-screen deployment. This study develops a UAV-assisted smoke-screen deployment framework that combines a multi-stage kinematic model, a time-varying line-of-sight (LOS) occlusion metric, and a hierarchical genetic algorithm (H-GA) for the joint optimization of UAV motion, release scheduling, detonation delay, and task assignment. Compared with fixed-parameter deployment, the proposed framework explicitly couples terminal engagement geometry with smoke cloud evolution and multi-UAV cooperation. Across five scenarios, cooperative five-UAV deployment against three simultaneous missiles achieves 14.44 s effective shielding, compared with 1.42 s under fixed-parameter deployment. Sensitivity analysis indicates strict operational constraints: heading errors of 2–3° reduce shielding by more than 50
This paper presents the development of a low-cost pyramidal solar sensor for space applications. The sensitive units of the developed sensor and their associated electronics were tested through several processes. By using a novel digital filtering algorithm, the signal quality was successfully enhanced. Furthermore, structural analysis proved aluminum is the best material choice. Finally, azimuth and elevation experiments and the comparison to other existing technologies confirmed that the proposed configuration provides a wide field-of-view, with a comparable accuracy to other commercial solar sensors while achieving approximately 60
This paper presents a rolling-horizon replanning framework for multi-UAV missions considering fuel constraints in dynamic environments. At each decision epoch, a re-optimization subproblem is solved to minimize the total travel distance of all agents (UAVs) while visiting remaining task locations, considering en-route refueling depots. A mixed-integer linear programming formulation of the subproblem is proposed to find the optimal solution. In addition, to enable real-time computation, we develop a heuristic-based solver framework: an extended sequential greedy algorithm provides a quick baseline, while a reinforcement learning (RL) approach using an attention model and long short-term memory aims to overcome the local optimality of the greedy method. A case study and numerical experiments demonstrate the effectiveness of the proposed framework and the performance of the solution methods.
Designing high-gain antennas within the constraints of the 1U CubeSat standard involves a trade-off: fixed-profile antennas can have limited bandwidth, whereas deployable mechanisms add mechanical complexity. This paper presents a deployment-free, frequency-reconfigurable antenna that integrates a motorized reflector within a cavity-backed CPW-fed architecture. The proposed Optimized Reflector-Integrated CPW-Fed Antenna (ORICPW-FA) is contained within a standard 1U structural frame and is intended as a communication payload for multi-unit CubeSats (e.g., 3U and 6U), while meeting the 6.5 mm rail-protrusion limit. Full-wave simulations indicate that reflector-height tuning extends the pattern-stable range from 6.0 to 8.2 GHz for the fixed-reflector reference configuration to 6.0−9.4 GHz for the tuned configuration, while maintaining MLD ≤ ±3° and SLL ≤ −12 dB. A laboratory prototype was evaluated using S₁₁ measurements and time-domain-gated radiation-pattern measurements at three representative tuning states: 7.0, 7.8, and 8.6 GHz. The measured peak realized gains were 8.41, 7.78, and 7.58 dBi; the measured patterns showed MLD values of 0°−3° and SLL values of −17.6, −18.9, and −14.6 dB, respectively. These measurements support the simulated pattern behavior at three representative tuning states; continuous full-band radiation-pattern validation and space-environment qualification are outside the scope of this laboratory study.
On-board models of turboshaft engines are fundamental to health monitoring and fault diagnosis, requiring high accuracy and strong generalization across the full flight envelope under real-time constraints. However, steady-state flight data are extremely scarce and unevenly distributed in practice, making it difficult to construct a generalized on-board steady-state model using conventional methods. To address this issue, this paper proposes a hybrid modeling approach that integrates component-level priors with flight data calibration. First, a mixture-of-experts (MoE) steady-state baseline model is built using component-level steady-state grid data to provide a global prior mapping across the entire envelope. Then, limited flight steady-state data are introduced to correct the systematic deviation between the baseline model and the actual onboard environment via a staged fine-tuning strategy. After calibration, the RMSEs for the three key parameters ( N_g , P_3 , T_45 ) are reduced by 98.18 P_3 , 2 T_45 ), the residual distribution of the MEL model exhibits clear and quantifiable shifts, whereas the conventional steady-state MoE shows almost no response, demonstrating the high sensitivity of the proposed model to abnormal conditions. The proposed method enables high-accuracy full-envelope modeling even under severely limited flight data, providing a reliable model foundation for onboard health monitoring of turboshaft engines.
This paper presents a nonlinear analysis of the longitudinal flight dynamics of an air-breathing hypersonic vehicle (ABHV) model, with particular emphasis on aero–propulsive coupling effects. Bifurcation and continuation methods are employed to systematically characterize the equilibrium structure and stability boundaries with respect to key control inputs, namely fuel-equivalence ratio and elevator deflection. The open-loop analysis reveals the presence of both static and oscillatory instabilities, including fold and Hopf bifurcations, together with strong modal interactions governing transitions between oscillatory and non-oscillatory behavior. Time-domain simulations are used to illustrate the associated dynamic responses in the vicinity of critical operating conditions. Based on these insights, a stability augmentation system (SAS) incorporating angle of attack and pitch rate feedback is developed and analyzed within the same continuation framework. The analysis is further extended to steady-level flight conditions to determine the trim envelope, evaluate its stability characteristics, and compute stabilizing gains along the attainable operating range. The results demonstrate that continuation-based techniques provide a systematic framework for characterizing nonlinear stability boundaries and supporting the design of stability augmentation systems for aero–propulsively coupled hypersonic vehicles.
Optimizing design parameters to enhance supersonic mixing is crucial for the design of supersonic flow devices, such as scramjet engines, chemical lasers, and supersonic ejectors. However, obtaining optimal parameters using reinforcement learning is challenging because data for the supersonic mixing process are expensive to generate through simulations or experiments. Herein, a deep policy gradient network with an adaptive noise-scaling framework is presented to improve sample efficiency in the optimization of pulsed injection. A sampling strategy with adaptive noise scaling is introduced to balance exploration and exploitation. A deep policy gradient network uses a predictor deep neural network to provide feedback to adjust the pulsed frequency, pulsed amplitude, and mean total pressure, thereby approaching the desired supersonic mixing performance. The single and multi-objective optimization are achieved. The optimization accuracy is verified through simulations of supersonic reactive flows in a chemical laser, with relative errors of the small-signal gain coefficient remaining below 5
Deep-sea maritime rescue of a distressed moving maritime vehicle and its survivors using heterogeneous fixed-wing unmanned aerial vehicles (FW-UAVs) requires multidirectional and synchronized resource deployment to establish a successful rescue. However, FW-UAVs have aerodynamic constraints, including the need for continuous forward motion and limited turning radii. These constraints result in complex cooperative task assignment and execution challenges, creating a multimodal optimization landscape in which traditional algorithms often stagnate. To address this, we propose a systematic framework that integrates an adaptive quantum-inspired genetic algorithm (AQIGA) with a vector field guidance model (VFGM). The AQIGA is a high-level task allocator in which a task is a specific approach direction (AD) for an FW-UAV to deploy its resources. It simulates a population of quantum probabilities to determine optimal directional assignments and uses a dynamic feedback loop based on entropy and diversity to regulate exploration and prevent local optima. At a lower level, a decentralized VFGM handles real-time task execution by guiding FW-UAVs along their assigned ADs. It generates collision-free trajectories for simultaneous multidirectional arrival at the rescue area while adhering to FW-UAV aerodynamic constraints. We tested our framework in simulated moving maritime rescue scenarios and compared it with conventional quantum-inspired benchmark algorithms. The AQIGA-VFGM demonstrated, on average, an improvement of 13.23
This numerical study examines hydrogen-enriched methane/air combustion under gas-turbine operating conditions (3 atm, 85
This paper investigates the finite-time (FT) tracking control problem of a type of unmanned helicopter (UH) system. The system under consideration involves system uncertainties, unknown external disturbances, input saturation and actuator faults. Fuzzy logic systems (FLSs) are employed to approximate the system uncertainties, and corresponding parameter adaptive laws are designed to estimate the associated weight vectors. Additionally, parameter adaptive laws are developed to estimate the upper bound of the composite disturbance, which includes the FLS approximation error, external disturbance, input saturation error, and unknown actuator bias fault. Based on dynamic surface control (DSC) technology and Lyapunov stability theory, adaptive FT fuzzy DSC strategies are proposed for both the position subsystem and the attitude subsystem. The proposed approach can guarantee the closed-loop UH system is semi-globally practically finite-time stable (SGPFS), while ensuring that all signals remain bounded. Finally, simulation results are provided to validate the tracking performance of UH under the proposed control strategy.
This study develops an AI-based pilot-assistance system for urban air mobility (UAM) that recommends alternative flight paths as operational conditions evolve—for example, as battery state-of-charge decreases or wind and gust severity increases. Trajectory generation uses the A* algorithm with a cost function that combines a distance term, favoring a shorter required time of arrival (RTA), and a risk term capturing static (building) and dynamic (wind, gust) hazards in the urban environment. The proposed AI pilot is a reinforcement-learning agent, trained with the deep deterministic policy gradient (DDPG) algorithm, that selects a continuous weighting factor λ to control the safety–efficiency trade-off by adjusting the relative weighting of the distance and risk terms based on the current operational context. Because battery and wind information are provided to the policy as continuous observations rather than discrete emergency flags, the resulting λ responds smoothly and monotonically to gradually degrading conditions, while still reacting decisively under abrupt emergencies. We validate the system through Monte Carlo benchmarks against rule-based and fixed- λ baselines, and through a controlled context-sweep case study. Evaluation is based on path length, risk exposure, and trajectory smoothness. These results demonstrate the feasibility of the AI pilot as a decision-support tool for both gradual context changes and emergency responses during UAM operations.
Compared with conventional solid rocket motors (SRM), solid rocket motors with pintle nozzle have unique advantages in the thrust regulation and mission maneuverability. But a phenomenon that thrust vector deviates from the axis of symmetry (i.e., thrust eccentricity) is always unavoidable, which can produce lateral force and affect the precise control of aircraft. In present study, the characteristics of thrust eccentricity are explored by establishing the numerical model of pintle nozzle, and the model accuracy is validated by the experiments of cold flow. Four potential factors that affect the thrust eccentricity, including the opening valve, the inlet pressure, the inlet angle and the pintle profile, are analyzed and evaluated by the eccentricity angle. The results show that the thrust eccentricity is caused by inherent asymmetry of internal flow field, which is caused by the side-inlet configuration of pintle nozzle. The eccentricity angle is significantly increased with both the valve opening and the inlet pressure, and it can be reduced by decreasing the inlet angle of fluid flow. Furthermore, an elliptical pintle profile can minimize the thrust eccentricity effectively, when compared with the linear profile. Across all investigated conditions, the overall thrust eccentricity angle ranges from − 0.70° to 2.65°. The results provide references for reducing the influences of eccentricity thrust, and can be used to design the pintle nozzle in solid rocket motor with variable thrust.
Reliable and cost-effective modeling of key components in rocket engines via deep learning has become a major focus of current research. Focus on the axial temperature field of a gas‑oxygen/gas‑methane igniter combustion chamber, this study presents an enhanced convolutional autoencoder (CAE) designed for dimensionality reduction and reconstruction. By balancing accuracy and dimensionality, the model achieves a compression ratio of up to 32:1. The proposed model exhibits robustness and generalization capability under noise interference and varying oxygen‑to‑fuel ratio conditions, with relative errors for both average temperature and the proportion of high‑temperature zones remaining below 6
This paper proposes a robust fault-tolerant control scheme to address the challenges of automatic carrier landing systems for aircraft under air-wake disturbances, actuator faults, and output constraints. First, a predefined-time prescribed performance function (PTPPF) is constructed to explicitly confine the tracking error within a predefined range while ensuring convergence within a predefined time, independent of initial conditions. Then, a predefined-time sliding mode controller (PTSMC) is designed to enhance robustness against nonlinear uncertainties and external disturbances. Furthermore, a prescribed-time extended state observer (PTESO) is designed to accurately estimate lumped disturbances and actuator faults within a prescribed time, enabling real-time compensation in the control loop. Lyapunov-based stability analysis proves that all closed-loop signals remain bounded and the tracking error converges within the predefined-time. Carrier landing simulation experiments, considering carrier air-wake, deck motion, and actuator faults, demonstrate that the proposed method significantly improves transient performance, disturbance rejection, and landing accuracy.
Collision avoidance (CA) is an important function for autonomous aircraft since midair collisions cause fatal risks in airspace. Thus, Detect-and-Avoid (DAA) systems and technologies have been widely studied for decades. Accordingly, standards for DAA systems have also been established, which contain sensor systems for detecting external risks and processing units for alerting risks and guiding appropriate maneuvers. As an effort to develop a DAA system, this study proposes detailed functional designs based on a DAA architecture defined in the standards, which consists of Track Processing, Track Alerting, and Guidance Processing functions. These functions are designed and implemented into a DAA system hardware. The designed system is validated through hierarchical test procedures. In addition, quantitative evaluations of a CA method used in the Guidance Processing are conducted and the results are compared with those of an existing CA method for the DAA system. The results show the proposed CA method is more computationally efficient than the existing method while containing competitive CA performance. As a result, this study suggests a DAA system design with an efficient CA processing method.
Actuator loss-of-effectiveness (LoE) in a single rotor reduces control authority and can destabilize fixed-gain quadrotor controllers. This paper investigates a bounded online gain-adaptation layer, implemented via reinforcement learning, for a standard cascaded proportional–integral–derivative (PID) quadrotor controller. The emphasis is not RL-based PID tuning per se, but a practical integration in which four decentralized agents update roll, pitch, yaw, and altitude PID gains in real time within preset bounds while the underlying mixer, outer-loop structure, and rigid-body model remain unchanged. A six-degree-of-freedom Newton–Euler quadrotor model with first-order motor dynamics is implemented in MATLAB/Simulink, and a Soft Actor-Critic (SAC) instantiation of the adaptation layer is evaluated under nominal flight and multiple transient and sustained single-rotor LoE profiles. For context, Twin Delayed Deep Deterministic Policy Gradient (TD3) and Proximal Policy Optimization (PPO) are evaluated under the same observation/action definitions, reward structure, gain limits, and 500-episode training budget; an additional 1000-episode PPO check is included to assess training-budget sensitivity. Offline-optimized fixed-gain PID baselines obtained using particle swarm optimization (PSO) and grey wolf optimization (GWO) are included to separate the benefit of online adaptation from static tuning and to address sensitivity to the selected metaheuristic baseline. Step responses and three-dimensional trajectory-tracking simulations show comparable nominal accuracy across methods but clear differences under degradation: online gain adaptation yields smaller post-fault deviations than fixed gains, while SAC provides the most consistent fault accommodation among the evaluated configurations. In the tested setup, these results support bounded online gain adaptation as a low-integration-cost way to improve LoE robustness without replacing a familiar PID-based quadrotor control architecture.
