Vision-and-Language Navigation (VLN) plays a critical role in tasks of embodied AI, particularly in unseen environments following natural language instructions. Recent advancements leverage large language models (LLMs) to improve the accuracy and generalizability of VLN systems by encoding image sequences as dense token representations. However, this tokenization approach incurs substantial computational overhead due to two key inefficiencies: 1) ego-centric camera views often include navigation-irrelevant re- gions (e.g., sky or distant backgrounds), and 2) high-frame-rate image sequences introduce temporal redundancy. To address these challenges, we propose Spatial-Temporal Efficient Visual Token Pruning (STEP-Nav), a unified frame- work that simultaneously prunes redundant visual tokens and fine-tunes VLN models to preserve navigation performance. In particular, STEP-Nav incorporates a distance- and content-aware token evaluation mechanism to remove irrelevant tokens at the spatial level, along with temporal level similarity-based filtering to reduce redundancy across sequential frames. To ensure pruning does not harm task performance, we introduce a distortion-aware fine-tuning strategy that aligns pruned-token representations with their full-token counterparts while maintaining navigation accuracy. Experiments on the R2R and RxR benchmarks using Navid-CE and NavGPT-2 as base models demonstrate that STEP-Nav preserves over 95% of the performance while reducing 66.7% of tokens, outperforming existing token pruning baselines.

Vehicle trajectory prediction is important for automated vehicles to understand driving scenarios. This paper proposes an encoder-decoder network-based parameterized transfer learning (EDN-PTL) model to predict vehicle trajectory. To improve trajectory prediction accuracy, the motion interaction between the target vehicle and the surrounding vehicles is considered, and a multidimensional spatiotemporal input expansion (MSIA) strategy is proposed to extend the feature dimensions. Additionally, global and local scale features, as well as long and short horizon features, are extracted and used for interactive vehicle trajectory prediction by a CNN and LSTM-based encoder-decoder network (CNN-LSTM-EDN). Moreover, the features extracted by CNN-LSTM-EDN are integrated using a stacked convolutional social pooling network (SCSPN). To enhance the environmental adaptability of the trajectory prediction model, a PTL strategy is proposed to enable transfer learning capabilities of EDN-PTL. Based on the PTL strategy, trajectory prediction accuracy is maintained even when applied to untrained environments. The proposed EDN-PTL model is validated on three types of publicly available naturalistic datasets and compared with several baselines and state-of-the-art (SOTA) methods. The validation results demonstrate that the proposed EDN-PTL achieves better prediction accuracy, robustness, and environmental adaptability compared to the baselines and SOTA methods.

Search time is an important metric for regional coverage searches conducted by unmanned clusters. This paper proposes an improved proximal policy optimization (IPPO) algorithm to decrease search time while ensuring the coverage rate for a heterogeneous cluster consisting of unmanned ground vehicles and unmanned aerial vehicles (UGV-UAV). The models of the UGV-UAV cluster, search scenario, and search constraints are first developed. Then, the IPPO algorithm is designed to simultaneously learn the cross-domain actions of UGVs and UAVs. The main advantage of the IPPO is that it can achieve the cross-domain cooperative learning (CDCL) mechanism, thus ensuring the collaboration consistency of the UGV-UAV cluster and enhancing search efficiency. To analyze the IPPO-based regional coverage search performance, three state-of-the-art (SOTA) methods are selected for comparison. The validation results demonstrate that the proposed method outperforms these SOTA methods in terms of both search time and coverage rate.

Human-machine cooperation can simultaneously leverage the strengths of both human drivers and machines, making it a promising solution for improving driving safety, comfort, and experience. This paper designs a heterogeneous synchronous reinforcement learning (HSRL)-based human-machine shared steering decision-making (HMSSDM) strategy for intelligent vehicles. First, the vehicle dynamics, which incorporate steering characteristics, are built to quantify human driver’s steering behavior. Additionally, the scenario-oriented driving constraints (SODCs) are established to demonstrate driving constraints from traffic participants, roadside obstacles, and traffic signs. Second, to enhance the rationality and reliability of steering behaviors, the human driver’s steering behavior is evaluated using a fuzzy logic strategy, and HSRL is proposed to simultaneously determine steering actions and allocate driving authority between the human driver and machine. The main advantage of HSRL is its ability to perform both continuous domain learning (CDL) and discrete domain learning (DDL) simultaneously. Finally, the proposed method is validated using a human and hardware-in-the-loop (HHiL) experimental platform. The comparison results demonstrate that the proposed method outperforms the comparison methods in terms of driving safety, comfort and experience.

Lane recognition is an important component of autonomous driving system and advanced driving assistance system (ADAS) for intelligent vehicles. In complex driving conditions, accurate and fast lane recognition is a challenging issue. In this paper, a vision-based geometric model (VBGM) is proposed for accurate and fast lane recognition in complex conditions. The framework of the VBGM includes an image preprocessing stage and a lane recognition stage. In the image preprocessing stage, the region of interest (ROI) is extracted from the original image, and the original image is transformed into an undistorted greyscale image. In the lane recognition stage, the lane contour is first extracted using the Roberts operator. Then, to accurately and quickly recognize the lane marking, a lane recognition coordinate system (LRCS) and a rotational LRCS (R-LRCS) are constructed. The distracting contours in abnormal regions are padded based on the LRCS using a contextual frames correlation (CFC) strategy, and the midpoints of the lane contour are identified based on the R-LRCS. Finally, an adaptive-order polynomial fitting model is built to fit the lane marking according to the midpoints in the LRCS. To evaluate the effectiveness of the proposed method, two state-of-the-art methods are selected for comparison. The comparative results indicate that the proposed method possesses a higher recognition rate and speed for lane recognition in complex conditions.

Converting natural language text into executable SQL queries significantly impacts the healthcare domain, specifically when applied to electronic medical records. Given that electronic medical records store extensive patient information in a relational multitable database, developing a Text-to-SQL parser would enable the correlation of intricate medical terminology through semantic parsing. A major challenge is designing a versatile Text2SQL parser applicable to new databases. A critical step towards this goal involves schema linking - accurately identifying references to previously unseen columns or tables during SQL creation. In response to these key challenges, we propose a novel framework—Linking Information Enhanced Text2SQL Parsing on Complex Electronic Medical Records (LI-EMRSQL). This model leverages the Poincaré distance metric detection procedure, utilizing induced relations to enhance the performance of pre-existing graph-based parsers and improve schema linkage. To enhance the generalizability of LI-EMRSQL, the detection process is completely unsupervised and does not necessitate additional parameters. On two conventional Text2SQL datasets and two EMRs Text2SQL datasets, the system delivers SOTA performance. Furthermore, notable enhancements in the model's comprehension and alignment of schemas are observed.

Abstract Distributed heterogeneous systems have been widely adopted in industrial applications by providing high scalability and performance while keeping complexity and energy consumption under control. However, along with the increase in the number of computing nodes, the energy consumption of distributed heterogeneous systems dramatically grows and is extremely hard to predict. Energy‐conscious task scheduling, which tries to assign appropriate priorities and processors to tasks such that the system energy requirement would be met, has received extensive attention in recent years. However, many approaches reduce energy consumption by extending the completion time. In this article, we focus on the scheduling problem of energy‐conscious tasks in distributed heterogeneous computing systems and provide an efficient approach to mitigate energy consumption while minimizing the overall makespan of parallel applications. First, based on the heterogeneous earliest finish time, a fitness function is proposed to balance the makespan and energy consumption. Then, by improving the crossover and mutation operations of the traditional genetic algorithm, we proposed an efficient scheduling approach named energy‐conscious genetic algorithm to optimize the priorities and processor allocation of tasks, with objectives of minimizing the system energy and makespan. Experiment results on real‐world applications and simulations with randomly generated task graphs demonstrate that the proposed approach outperforms in energy‐saving and makespan reducing.

The energy utilization efficiency of autonomous electric vehicles is seriously affected by the longitudinal motion control performance. However, the longitudinal motion control is constrained by the driving scene. This article proposes an energy-saving optimization and control (ESOC) method to improve the energy utilization efficiency of autonomous electric vehicles. In ESOC, the constraints from the driving scene are thoroughly considered, and the autonomous driving scene constraints are mapped to the vehicle dynamics and control domain. On this basis, the efficiency self-searching method and the multiconstraint energy-saving control strategy are designed. The main ideology of the proposed ESOC is that the energy utilization efficiency of an autonomous electric vehicle can be improved by optimizing and controlling the operation point distribution of the powertrain efficiency. The experimental results demonstrate that the operation point distribution of the autonomous electric vehicle's powertrain efficiency can be well optimized by the proposed ESOC, and the energy consumption results indicate that the proposed ESOC outperforms the state-of-the-art methods.

Unmanned aerial vehicles (UAVs) have been widely applied in civilian and military applications due to their high autonomy and strong adaptability. Although UAVs can achieve effective cost reduction and flexibility enhancement in the development of large-scale systems, they result in a serious path planning and task allocation problem. Coverage path planning, which tries to seek flight paths to cover all of regions of interest, is one of the key technologies in achieving autonomous driving of UAVs and difficult to obtain optimal solutions because of its NP-Hard computational complexity. In this paper, we study the coverage path planning problem of autonomous heterogeneous UAVs on a bounded number of regions. First, with models of separated regions and heterogeneous UAVs, we propose an exact formulation based on mixed integer linear programming to fully search the solution space and produce optimal flight paths for autonomous UAVs. Then, inspired from density-based clustering methods, we design an original clustering-based algorithm to classify regions into clusters and obtain approximate optimal point-to-point paths for UAVs such that coverage tasks would be carried out correctly and efficiently. Experiments with randomly generated regions are conducted to demonstrate the efficiency and effectiveness of the proposed approach.

Heterogeneous multiprocessor platforms have been widely adopted as an efficient approach to providing high instruction throughput while keeping power and complexity under control. Although this approach can achieve improved performance for large-scale real-time systems, it results in a complex task scheduling problem. All tasks should be scheduled according to a proper strategy such that their deadlines will be met even in the worst case situations. In this work, we study the non-preemptive scheduling problem of periodic tasks with data dependency upon heterogeneous multiprocessor platforms. We first analyze the space, time and precedence constraints of tasks, and propose an exact formulation to determine the schedulability of tasks. Then, inspired from the Heterogeneous Earliest Finish Time (HEFT) algorithm, we present a list-based scheduling heuristic to schedule the jobs generated by the periodic tasks and minimize the jobs' finish time. The proposed approach is efficient and can help in guiding the design of heterogeneous multiprocessor systems.

Integrated modular design has been widely adopted as an approach to facilitating the development process of large-scale real-time systems. Although this approach can achieve enhanced design reuse and reduced time consumption, it results in a complex task allocation and scheduling problem. All tasks should be integrated into a shared platform according to a proper schedule, such that their deadlines will be met even under the worst case situations. In this paper, we study the allocation and scheduling problem of strictly periodic tasks in multi-core real-time systems. We first propose a necessary and sufficient condition to determine whether a new task is schedulable on a processor without changing the start times of the existing tasks. Then, based on the condition derived previously, we present a task assignment algorithm, which not only provides valid start times and processor allocations for all tasks, but also obtains the minimum number of processors required by the system. Finally, simulation experiments with randomly generated task sets are conducted to show the high efficiency and reliability of the proposed approach.

With the advent of embedded and mobile computing techniques, software systems are increasingly operated in open and dynamic environments. Such systems desire self-adaptive capabilities. This paper proposes an agent-oriented approach to the modeling and simulation of distributed adaptive systems. The approach enables to construct structural and behavioral models at a high abstraction level, and to validate the models at design stage through model simulation. The modeling language facilitates devising decentralized adaption logic at both component and system levels. Simulation demonstrates how a system dynamically re-organizes in response to the changes in the operating context. Case studies show that our approach can support various self-adaptation mechanisms with which a multi-agent system adjusts to internal failures, unexpected environment, or user's changing requirements.

Video data compression is one the core technologies of video conference, which aims to enable significantly high performance data coding. In order to achieve this, a robust rate-distortion optimization (RDO) technique is needed to be employed to select the best coding mode This paper presents optimization solution about inter prediction mode decision algorithm of H.264/AVC for video conference under the HLA/RTI framework. Our objective is to reduce the computational complexity of the CODEC (Coder & Decoder) with insignificant rate-distortion performance degradation.