In human-machine (autonomous vehicles and robots) interaction scenarios, pedestrians often appear in groups. Pedestrian groups provide richer information compared to individuals, which helps address occlusion problems in pedestrian-machine interactions. However, the randomness and spatiotemporal complexity of pedestrian activity make pedestrian group activity recognition (PGAR) a highly challenging task. This article provides a detailed description of the PGAR task. For the first time, a definition of pedestrian group and activity for autonomous vehicles and robots is provided. Existing datasets and methods are systematically summarized. Furthermore, the unique challenges and trends in PGAR for autonomous vehicles and robots are outlined. Although some related surveys have been published, there has not yet been a survey specifically focused on PGAR in autonomous driving and robotics scenarios. Therefore, the goal of this article is to narrow the gap in this topic and provide a comprehensive reference for researchers in this field.

Amid the rise of the low-altitude economy, hybrid electric unmanned aerial vehicles (HEUAVs) have emerged as a promising solution to mitigate energy scarcity and advance carbon neutrality. Owing to their high power-to-mass ratio, turboshaft engines are widely adopted as the primary power source for HEUAVs. Modern energy management strategies (EMSs) optimize power distribution between the turboshaft engine-generator pack (TGP) and the battery to enhance fuel economy. Besides energy efficiency, turbine durability necessitates incorporating the turboshaft engine's exhaust gas temperature (EGT) into EMS design, as prolonged over-temperature exposure risks turbine damage. The inherent nonlinearity of the TGP alone presents significant challenges, and the integration of nonlinear EGT dynamics further compounds the complexity, making the EMS a highly nonconvex optimization problem with severe computational demands. To address these challenges, this work proposes an efficient EMS for HEUAVs that jointly optimizes fuel economy and EGT regulation. First, the multi-objective EMS is formulated as a quadratic constrained quadratic programming problem under a model predictive control framework. Second, this problem is recast as a semidefinite programming (SDP) problem, where new-induced nonconvexities are consolidated into a rank-one constraint. Third, leveraging the sparsity of constraint matrices, the rank-one constraint is decomposed into smaller submatrix constraints to reduce computational complexity. An iterative rank minimization method then converts the nonconvex SDP into a sequence of convex subproblems, enabling efficient solutions. Simulations using real-world flight data and hardware-in-loop experiments validate the strategy's effectiveness, demonstrating an 8.8-15.6 % reduction in fuel consumption, and a 5.5-15.0 % improvement in EGT regulation, compared to the baseline method. Furthermore, hardware-in-loop results demonstrate the strategy's real-time applicability, underscoring its potential for scalable deployment in HEUAV operations.

It is widely acknowledged that heavy-lift electric vertical take-off and landing aircraft (eVTOLs) are considered to be pivotal components of the low-altitude economy, with the potential to facilitate high-payload transportation within complex urban low-altitude environments. To ensure safe operation in such a complex environment, accurate trajectory tracking during autonomous flight becomes essential. However, significant challenges arise from pronounced nonlinear dynamics, large inertia, and strong coupling effects. These factors degrade the performance of trajectory tracking controllers. To address these issues, a novel tracking method with a dual-loop framework is proposed. First, a feasible-region-based prescribed performance control (PPC) strategy is developed in the outer loop. Position tracking errors are strictly constrained within predefined time-varying envelopes to guarantee trajectory safety. Second, a physics-data integrated linear parameter-varying model predictive control (LPV-MPC) scheme is employed for the inner attitude loop. Angular rates are utilized as scheduling variables to capture rate-dependent nonlinearities and compensate for model uncertainties. Simulation and real-world flight experiments are conducted to validate the proposed method. The results demonstrate superior tracking accuracy, improved disturbance rejection, and enhanced safety performance compared with conventional benchmark approaches. And as demonstrated by the quantitative robustness analysis, a zero violation rate of the prescribed performance constraints is maintained even under a significant mass perturbation of 85.7%.

In steer-by-wire (SBW) systems, the steering performance is highly dependent on reliable steering angle measurements and steering torque output. Steering motor faults and steering angle sensor faults can both distort the steering response and may compromise vehicle safety. Fault-tolerant control (FTC) is therefore required to maintain steering functionality, and its effectiveness relies on accurate fault estimation. However, these faults can produce similar effects on system outputs under the same operating conditions, making single-fault estimation indistinguishable and degrading estimation accuracy. To address this issue, an integrated FTC scheme for SBW systems based on steering motor and steering angle sensor fault estimation is proposed. First, an improved fast adaptive unknown input observer (IFAUIO) is developed for simultaneous estimation of steering motor faults and steering-angle sensor faults. A differencing operation is introduced into the fast estimation law to suppress steady state estimation errors caused by bias faults, thereby improving fault estimation accuracy. Second, based on the estimated fault information, a high-order control barrier function-based model predictive control (HOCBF-MPC) fault-tolerant controller is developed to maintain steering angle tracking performance and coordinated steering-driving control under fault conditions. Finally, simulation and experimental results demonstrate that the IFAUIO improves the estimation accuracy of steering motor faults and steering angle sensor faults by 20.1 and 80.9%, respectively, compared to the fast adaptive unknown input observer (FAUIO). In addition, the proposed HOCBF-MPC improves steering angle tracking accuracy by 17.0% compared to the conventional MPC-based fault-tolerant controller.

Mode shift is a special mechanism for a power-split hybrid electric vehicle (HEV) to realise electrically variable transmission, but the sudden change of equivalent inertia caused by topological configuration recombination during mode shift induces a significant torque shock. Therefore, a smooth transient process, among other concerns, typically associated with this category of vehicles, is of great importance. The present research aims to introduce a novel control strategy to manage the dynamic torque of multiple power sources and therefore improve ride comfort. To this end, a dynamic model of the objective power-split HEV is first built. To resolve the contention between vehicle jerk and clutch friction loss, a model predictive control (MPC) combined with control allocation (CA) is then designed for the clutch-engaged phase. To reduce the torque fluctuation caused by the inertia torques of multiple power sources, a dynamic compensation control strategy (DCCS) that coordinates motor-generator torque to compensate for the transition torque is proposed for the brake-disengaged phase. Finally, the proposed control strategy is validated by simulation and bench test, and results show great potential in reducing shift duration, torque variation, vehicle jerk and friction loss (the simulation results show decreases of 22%, 39%, 83% and 53%, and the experimental results show decreases of 21%, 74%, 77%, and 59%, respectively), thereby improving shift quality.

Autonomous vehicles are further improved with the development of in-vehicle networks and X-by-wire technologies. Due to the increasing integration of vehicle systems and the growing complexity of control links, time delays in information transmission and execution have become increasingly prominent. Because of late responses to control inputs, autonomous vehicles have deterioration in terms of path following accuracy and control stability, which poses a threat to driving safety. To solve this problem, an adaptive model predictive path following control strategy is presented in this paper. Firstly, a real-time delay estimator is designed based on historical feedback information processing and the gradient descent algorithm. Secondly, based on the estimated delay, a delay-augmented model with variable dimension is proposed for adaptive control to mitigate the impact of time-varying delay. Meanwhile, a tube-based model predictive control method is modified for the proposed model to deal with uncertainties generated by time-varying delays. Finally, the proposed control strategy is verified by simulations and experiments in multiple cases. The results show that the proposed strategy achieved a smaller lateral offset than that in the ordinary model predictive control method, leading to a 52.86% improvement in accuracy under the existence of time-varying delays.

The accurate steering angle is the foundation of steering closed-loop control of steer-by-wire (SBW) system and vehicle driving safety. Since the SBW system relies on sensor to obtain the steering angle signal, the high-performance angle fault-tolerance estimation is crucial when the sensor occurs offset or failure faults. However, the sensor faults signal and the system nonlinear characteristics that difficult to model will reduce the angle estimation performance in the form of noises. This leads the accurate and stable angle fault-tolerant estimation to remaining challenging. To solve this issue, this paper proposes an adaptive angle fault-tolerance estimation for SBW system using hybridizing physical and data-driven model. Firstly, a deep neural network is designed and trained to capture the vehicle system nonlinear characteristic and establish data-driven model, as well as obtain additional angle information of SBW system. Secondly, an adaptive unscented Kalman filter (UKF) combining a confidence level mechanism is constructed. Based on the density of training data, the confidence level mechanism can dynamically adjust the covariance of UKF to balance the degree of dependence between the system physical model and data-driven model, which can improve the interpretability of the data-driven model and the accuracy of angle estimation. Then, to enhance the reliability of angle estimation, an expansion fault-tolerant mechanism is constructed to suppress the influence of abnormal measurement values when the angle sensor fails. Finally, the proposed method is verified using a vehicle equipped with SBW system. The experiment results demonstrate that the proposed method can accurately estimate steering angle and improve the fault-tolerant performance. The angle estimation application experiment further verifies the feasibility of the proposed method in the steering closed-loop control of SBW system.

Hybrid electric energy systems (HEES), as efficient and clean solutions, have become increasingly popular and are widely used in ground vehicles and aircraft applications. Parameter matching (PM) is critical to HEES, as it fundamentally dictates the achievable levels of reliability, performance, and economy. Current PM methods inherit from traditional single-power-source powertrains, focusing on power and energy boundaries, but neglecting HEES dynamic characteristics. This induces power-supply mismatches under high-dynamic conditions such as vehicle acceleration or aircraft attitude adjustment, causing bus voltage and engine speed fluctuations, thereby degrading powertrain power output performance. To address this, a dynamic-response-aware PM via a two-stage multi-fidelity NSGA-II is proposed. First, a frequency-domain hierarchical power allocation mechanism is introduced, breaking static limitations. The mechanism utilizes component frequency response differences to overcome single time-scale constraints, achieving precise synergy between system-level management and component-level specifications. Second, dynamic equivalent models are developed, replacing ideal-source assumptions. The models accurately capture transient nonlinear characteristics, correcting static deviations for high-fidelity optimization. Third, the two-stage multi-fidelity NSGA-II framework is established, enhancing optimization efficiency. The framework combines global coarse screening with local refined search, ensuring engineering accuracy while solving the trade-off between computational cost and optimization quality in high-dimensional problems. As a result, with dynamic models and frequency-domain power allocation, the proposed method reduces power errors by 6.8% and 7.7% and decreases bus-level RMS and peak errors by 7.3% and 11.5%. Lower battery lifecycle costs by 12.4% are achieved, while the optimized framework cuts runtime and computation costs by 46.8% and 40.3%.

Reliable power-following capability is essential for series hybrid electric vehicles (SHEVs), yet practical discrepancies between actual load-side demand and source-side generation can produce source-load power mismatch, thereby degrading charge-sustaining performance. Additionally, the ability of the power coordination strategy to reconcile transient maneuverability with steady-state fuel economy is also critical for unlocking the full potential of SHEVs. To address these challenges, this paper proposes a power-demand-informed robust power coordination strategy for SHEVs with prescribed-performance constraints. First, source-load power mismatch is formulated as an equivalent battery-port residual and mapped into an additive disturbance, which connects practical unmodeled powertrain uncertainties with the control-oriented model. Second, a power-demand-informed factor is constructed to adaptively blend economy-oriented and dynamic oriented engine speeds, thereby accommodating diverse driving demands. Third, an extended state observer is designed to estimate the disturbance. Then, its bounded estimation error is derived and incorporated as a robustness margin into the transformation from prescribed-performance requirements to battery power constraints. The resulting nonlinear model predictive control problem is efficiently solved with constrained iterative linear quadratic regulator. Simulation and hardware-in-the-loop (HIL) tests are conducted to validate the performance of the proposed strategy. Compared with the baseline strategies, the proposed strategy achieves a favorable compromise among fuel economy, engine speed stability, and SoC robustness across both testing scenarios. The root mean square error of SoC is reduced by 97.778% and 97.297%, respectively, demonstrating its robustness against source-load power mismatch. The HIL test results further confirm the practical application potential of the strategy.

Hybrid electric vehicles (HEVs) have emerged as a promising solution for achieving carbon neutrality and overcoming energy-related challenges. The efficacy of HEVs, in part, is contingent upon the implementation of effective energy management strategies (EMSs). Recently, on the one hand, most EMSs have primarily emphasized fuel economy while overlooking the potential consequences of frequent and high-power usage of electronic components, which may lead to motor overheating and the degradation of the motor's maximum capacity. On the other hand, benefiting from future driving conditions in a finite horizon, prediction-based EMSs can proactively make some adjustments to energy allocation in advance, garnering significant attention from numerous researchers. However, inherent prediction uncertainties will mislead predictive EMSs to make overconfident decisions and ultimately introduce risks to thermal dynamics. Targeting the issues above, this article presents a min-max game-based generator temperature-sensitive EMS for series HEVs (SHEVs). First, a multiobjective energy management problem with a comprehensive consideration of fuel economy and generator temperature is formulated. Then, a min-max game framework for robust decision-making is structured to deal with the potential adverse effects of prediction uncertainty. Finally, to improve computational efficiency, a closed-form solution is proposed to solve the Stackelberg equilibrium pair for the min-max game, which avoids performing optimization in each iteration. Simulation and hardware-in-the-loop test results demonstrate that the proposed strategy enhances fuel economy while effectively lowering generator temperature.

Trajectory planning for unmanned ground vehicles (UGVs) involves creating a feasible path that includes a speed profile, essential for ensuring driving safety. Compared to structured roads, off-road environments frequently feature rough terrain, which possibly lead to significant rolling of UGVs, thereby compromising driving safety. To address this problem, an unequal probability sampling rapidly exploring random tree (UP-RRT*) method is proposed for UGVs to navigate rough terrains effectively. Initially, a novel flatness extraction method is introduced to refine the sampling process. This method adjusts the sampling range and probability by correlating the capability of UGVs with terrain characteristics, significantly enhancing the safety of autonomous driving in off-road environments and accelerating the convergence of the RRT* algorithm. Subsequently, a cost function designed for off-road environments is developed to optimize the structure of the random tree. Once the random tree extends to the designated goal point, it is smoothed using the cubic spline method to generate a preliminary trajectory. Lastly, the trajectory’s speed profile is dynamically adjusted by constraining the attitude of the UGVs in the prospective planned trajectory. The proposed method is verified by experiments in rough terrain with an actual vehicle. Results show that the absolute average value of the roll angle and pitch angle of the UP-RRT* method is decreased by 69.3

The steer-by-wire (SBW) system provides flexibility in vehicle design layout and steering ratio customization. However, the mechanical transmission components in the SBW system can cause a dead-zone effect, which degrades steering angle tracking performance and consequently affects steering maneuverability. To address this issue, a generalized regression neural network-based adaptive nonsingular fixed-time sliding mode control (GRNN-ANFSMC) method is proposed for the SBW system of vehicles. First, a dynamic model of the steering actuator considering input dead zone is established. Next, an improved fixed-time stable system is designed to achieve a faster convergence rate and a smoother convergence process. Building on this, a segmented sliding mode surface is designed to prevent singularity issues, and a reaching law with adaptive parameters is constructed to prevent chattering, which can enhance the dynamic response of the SBW system. Meanwhile, GRNN is introduced, and its weights are updated in real time by the designed adaptive law, which can effectively approximate and compensate for the dead zone of the SBW system within a fixed time. Finally, the effectiveness of the proposed method is validated by the SBW-equipped vehicle. The experimental results demonstrate that compared to other fixed-time sliding mode control (SMC), the proposed method can effectively reduce the angle tracking error and improve the dynamic response and robustness of the SBW system with a time-varying nonlinear dead zone.

Heavy-lift electric vertical take-off and landing (eVTOL) aircraft play a vital role in advancing the low-altitude economy by enabling the transportation of high payloads in dense urban environments. However, autonomous navigation in such environments presents significant challenges for heavy-lift eVTOLs due to the large moments of inertia and limited actuator capabilities, which hinder agile maneuvers and effective obstacle avoidance in confined spaces. Motivated by the mentioned issue, a deep reinforcement learning (DRL) path planning method that explicitly incorporates a data-driven dynamic flight characteristic model (DFCM) is introduced to plan safe and feasible flight paths for heavy-lift eVTOLs. First, the DFCM is formulated to fully consider the flight control performance of the heavy-lift eVTOL. And then, utilizing the built DFCM, the real-time feasible attitude domain can be obtained to constrain the autonomous motion attitude of the heavy-lift eVTOL. Third, the path planning method considering motion attitude constraints is established through the DRL process. The accuracy of the proposed model has been verified using real flight data, with the corresponding correlation coefficients amounting to 0.9723. The experimental planning results show that the proposed method generate safe and executable trajectories within the feasible attitude domain.

In high-speed and complex highway scenarios, occlusions and rapid vehicle speeds significantly reduce available reaction time. Current autonomous driving systems primarily address visible risks, making it difficult to detect and respond to latent risks in occluded situations or scenarios that require reasoning. This paper presents TRACE-MPC (Triggered risk Abduction and Compliance-coupled Model Predictive Control), an ontology-driven risk reasoning and planning framework based on unexpected-behavior analysis that identifies unsafe, non-compliant, or unreasonable behaviors and analyzes the hidden risks they imply. We construct a seven-layer scene ontology as the foundation for reasoning. The ontology is supported by prior knowledge and traffic regulations, forming a rule base that defines and judges unexpected behaviors through interaction logic among scene elements. Potential-risk analysis uses harmful-interaction relations between traffic participants as mediating links for risk propagation. The reasoning outputs feed an MPC module to enable safe and smooth trajectory planning. Experiments in representative complex scenarios show that TRACE-MPC improves minimum time-to-collision by 172% under occlusion and avoids collision in a pedestrian-occlusion case where baselines impact at 32–35 km/h, while also improving post-merge spacing and reducing braking intensity during ramp merges, demonstrating reliable latent-risk identification and decision support.

Path planning plays an important role in autonomous flight in the complex urban environment to achieve goods delivery tasks. Besides dense obstacles, the wall effect near tall buildings brings great risks to the unmanned delivery aerial vehicle (UDAV). The objective of this work is to mitigate the risks and improve the safety and reliability of the UDAV. Motivated by this objective, a risk-aware obstacle avoidance path planning method for UDAV is proposed. First, the model of ducted fan-type UDAV is introduced. By analyzing its driving environment, different kinds of risks, including the wall effect, are constructed as an urban risk map, which reduces computational complexity by mapping them from three-dimensional space to two-dimensional planes. Then, a risk-aware improved artificial potential field (APF) path planning method is proposed, with a precognitive architecture, to plan a feasible path avoiding risks. The infeasible problems of the traditional APF method are solved by adding pre-decision-making instruction and continuous actions. Finally, scenarios are built to verify the effectiveness of the proposed method. Results show that the proposed method can plan a safe path in a complex urban environment. Compared to the planned path proposed by the popular particle swarm optimization method and the APF method, the risk index is decreased by 16.25% and 14.10%, respectively.

Current trajectory planning approaches mainly concentrate on geometric optimization and do not take vehicle nonlinear dynamics into full consideration. Infeasible lane-change trajectories may cause dynamic instability in critical situations, especially under low road frictions. This paper presents a dynamic-stability-aware lane-change planning method (DSA-LCP) formulated as an optimal control problem for automated vehicles. Trajectory smoothness, geometrical safety, and handling dynamics are integrated into the formulations of this planning method. Specifically, the road friction coefficient, which has crucial impacts on handling limitations, is deeply analyzed to within road adhesion capacity. Various test results including various road conditions and vehicle speeds are carried out, and the results validate the efficiency of the proposed DSA-LCP.

With advances in electric drive technology, electric tracked vehicles (ETVs) have emerged as a promising solution for high-mobility ground vehicles. However, under high-speed steering conditions, the equivalent motor load inertia varies significantly, introducing strong nonlinear and time-varying characteristics into the ETV that may induce lateral instability and even rollover. To address this issue, a novel augmented deep Koopman operator-based model predictive control (ADK-MPC) method is proposed. First, a high-order sliding-mode (HOSM) observer is designed to estimate the lumped load disturbances associated with the time-varying equivalent motor load inertia. Then, the estimated disturbances are introduced as an augmented state into the DK operator to construct a data-driven augmented model. The proposed model transforms the nonlinear dynamics into a lifted linear time-invariant representation in the augmented-state space while capturing the dominant nonlinear characteristics. Based on the ADK model, an ADK-MPC controller is developed to convert the nonlinear optimization problem into a quadratic programming problem, thereby improving steering stability and reducing computational complexity. Simulation results under steering conditions indicate that the proposed method achieves better yaw rate tracking and lower computational cost than nonlinear MPC. The yaw rate tracking error is reduced by 45.5%, while the average solving time is shortened by 11.7%.

Large-scale multirotor unmanned delivery aerial vehicles (UDAVs) are increasingly used in low-altitude transport due to their high payload capacity and maneuverability. Fault detection is crucial for UDAV flight safety, as it identifies potential failures before they escalate, drawing widespread attention as a key technology. Despite its promising prospects, this technology faces the challenges of insufficient extraction of spatiotemporal features from flight data and difficulty in capturing fault information from the extracted features. To address these issues, an enhanced spatiotemporal feature extraction and fusion framework combined with a multi-head attention mechanism fault classifier (ESTEF-MH) is proposed for UDAV fault detection. First, the feature extraction and fusion framework introduces bidirectional long short-term memory and 2D convolutional neural network to extract features from the temporal and spatial dimensions. It then merges these features, enabling comprehensive capture of flight information. Second, a decoupled coordinate attention mechanism is used to enhance the spatiotemporal feature representation. Finally, a multi-head attention-based fault classifier is designed for efficient fault classification. The method is tested on four actuator faults and the normal conditions, with three ablation experiments and comparisons to classic methods. Furthermore, actual flight test was conducted with results showing 84.44% accuracy, validating the reliability and adaptability of our method in real-world operational environments. Compared to other methods, this approach shows superior performance in terms of key metrics, offering a reliable guarantee for flight safety.

Power coordination strategy (PCS) plays a crucial role in realizing the desired performance of series hybrid electric vehicles (SHEVs). The inherent uncertainty of driving scenarios is considered one of the inevitable challenges that PCS need to deal with, accompanied by varying driving demands across diverse conditions. Enhancing the overall performance of SHEVs under frequently changing driving conditions represents a critical obstacle faced in strategic design. To tackle this issue, this paper proposes a driving intention awareness-based multi-objective robust coordinated control strategy for SHEVs. In contrast to conventional strategies that overlook the dynamic response characteristics of components, an integrated control model for series hybrid powertrain aimed at coordinated control is constructed. To promote the strategy’s capability to cope with uncertain operating conditions, a driving intention representation method is devised to quantify the uncertainty set of driving demands. Furthermore, a driving intention awareness-based objective transfer function is designed to achieve adaptive alignment of performance priorities with varying driving demands. The developed multi-objective robust coordinated control problem is efficiently solved with distributed optimization algorithm. Finally, simulation and hardware-in-the-loop (HIL) test are conducted to validate the performance of the strategy. With other performance metrics remained similar, the proposed strategy demonstrates 3.22% and 3.41% improvement in fuel economy compared to power-following strategy under two testing scenarios, while also exhibiting superior property in reducing fluctuations of engine speed. Additionally, the proposed strategy’s capability for objective transfer and robustness across diverse driving conditions is also verified. The results in HIL test further confirm the application potential of the strategy.

The high-performance tracking control of steering angle for steer-by-wire (SBW) system is foundation of vehicle driving safety. However, the variable steering load will reduce the tracking performance in the form of system disturbance, thereby causing steering hysteresis, which is uncertain, nonlinear, difficult to be modeled precisely and addressed effectively. This leads the accurate and stable steering angle tracking control to remaining challenging. In order to solve this issue, this article proposes an adaptive pseudoinverse fuzzy control method using nonlinear-quantization neural network for SBW system. First, to dynamically describe the uncertain steering hysteresis of SBW system online, a pseudoinverse compensator is constructed using fuzzy nonlinear-quantization cerebellar model articulation neural network, which avoids the complex dynamics modeling and inverse calculation. The fuzzy logic system is introduced to nonlinearly quantize the neural network input to improve the compensation accuracy of pseudoinverse compensator without increasing the computational burden. Then, to reduce the tracking error caused by the system disturbance from the variable steering load, an adaptive fuzzy controller with adjustable fuzzy mapping is proposed to enhance the antidisturbance control ability in the tracking process of steering angle. Finally, the proposed control method is verified by a vehicle equipped with SBW system. Experimental results show that, the proposed control method can effectively reduce steering hysteresis caused by variable steering load, and improve the tracking accuracy and stability of steering angle of the SBW system.