Multiple pedestrian tracking is crucial for enhancing safety and efficiency in intelligent transport and autonomous driving systems by predicting movements and enabling adaptive decision-making in dynamic environments. It optimizes traffic flow, facilitates human interaction, and ensures compliance with regulations. However, it faces the challenge of tracking pedestrians in the presence of occlusion. Existing methods overlook effects caused by abnormal detections during partial occlusion. Subsequently, these abnormal detections can lead to inaccurate motion estimation, unreliable appearance features, and unfair association. To address these issues, we propose an adaptive occlusion-aware multiple pedestrian tracker, OccluTrack, to mitigate the effects caused by partial occlusion. Specifically, we first introduce a plug-and-play abnormal motion suppression mechanism into the Kalman Filter to adaptively detect and suppress outlier motions caused by partial occlusion. Second, we develop a pose-guided re-identification (Re-ID) module to extract discriminative part features for partially occluded pedestrians. Last, we develop a new occlusion-aware association method towards fair Intersection over Union (IoU) and appearance embedding distance measurement for occluded pedestrians. Extensive evaluation results demonstrate that our method outperforms state-of-the-art methods on MOTChallenge and DanceTrack datasets. Particularly, the performance improvements on IDF1 and ID Switches, as well as visualized results, demonstrate the effectiveness of our method in multiple pedestrian tracking.

Existing human object interaction (HOI) detection methods have introduced zero-shot learning techniques to recognize unseen interactions, but they still have limitations in understanding context information and comprehensive reasoning. To overcome these limitations, we propose a novel HOI learning framework, ContextHOI, which serves as an effective contextual HOI detector to enhance contextual understanding and zero-shot reasoning ability. The main contributions of the proposed ContextHOI are a novel context-mining decoder and a powerful interaction reasoning large language model (LLM). The context-mining decoder aims to extract linguistic contextual information from a pre-trained vision-language model. Based on the extracted context information, the proposed interaction reasoning LLM further enhances the zero-shot reasoning ability by leveraging rich linguistic knowledge. Extensive evaluation demonstrates that our proposed framework outperforms existing zero-shot methods on the HICO-DET and SWIG-HOI datasets, as high as 19.34% mAP on unseen interaction can be achieved.

Visual Multiple Object Tracking (MOT) typically utilizes appearance and motion clues for associations. However, these features may be limited under certain challenging scenarios, such as appearance ambiguity and frequent occlusions. In this paper, we introduce a novel deep RF-affinity neural network (DRFAN) that enhances visual tracking with the aid of a passive wireless positioning device, Radio Frequency Identification (RFID). DRFAN aims to solve object tracking by introducing a new concept of a "candidate trajectory" to indicate target movement. This approach fundamentally deviates from existing fusion methods that rely on known visual tracks. Instead, DRFAN exclusively uses detection bounding boxes and RFID signals. The proposed method overcomes the limitations of visual tracking by swiftly resuming correct tracking whenever a failure occurs. This is the first time using signals from low-cost passive RFID tags to achieve image-level localization, and a discriminative neural network is designed specifically for RFID-assisted visual association. Our experimental results validate the robustness and applicability of the proposed approach.

Automated Guided Vehicles AGVs play a vital role in enhancing productivity and efficiency within factory environments. However, their safe and effective operation heavily relies on the ability to navigate through complex spaces while avoiding obstacles. The significance of obstacle avoidance in AGV systems is emphasized, considering its impact on ensuring smooth material flow, minimizing collision risks, and optimizing production processes. The existing state of obstacle avoidance applications in factory settings reveals certain limitations and challenges. Current research and industrial implementations often rely on rule-based approaches or predefined paths, which may not adequately adapt to dynamic environments or unexpected obstacles. Additionally, some methods lack the ability to handle diverse obstacle types or efficiently plan optimal paths, leading to sub-optimal navigation or reduced throughput. In response to these challenges, this study proposes a novel approach for dynamic obstacle avoidance of AGVs based on deep reinforcement learning. By leveraging the Deep Deterministic Policy Gradient (DDPG) model, the AGV learns to make real-time decisions and navigate through dynamic obstacles effectively. The integration of deep neural networks with the actor-critic framework enables the AGV to learn and adapt optimal policies for obstacle avoidance in real-time, overcoming the limitations of rule-based methods. Simulation experiments are conducted to validate the performance and feasibility of the proposed approach. The results demonstrate that the DDPG-based method allows the AGV to successfully navigate through both dynamic and static obstacles in a dynamic environment, improving safety and efficiency in intelligent manufacturing applications.

Permanent magnet vernier machines (PMVMs) are recognized as the most promising candidates for high-torque direct-drive applications. However, their market penetration is hindered by the low power factor. This paper aims to address the low-power-factor challenge, by employing the concept of low-coupling winding. As compared with two conventional PMVMs, the proposed PMVM with low-coupling winding exhibits a substantially improved power factor. To objectively evaluate the potential of the proposed PMVM, a comprehensive comparison between the proposed PMVM and the conventional permanent magnet synchronous machine (PMSM) is conducted. The no-load back-EMF, output torque, power factor, core losses, and efficiency are compared under various operating conditions, encompassing the entire torque-speed range. The strengths and weaknesses of the PMVM are analyzed, and the most promising application scenarios are suggested. Finally, a prototype of the proposed PMVM is fabricated to validate the analysis and comparison results.

The inverter nonlinearity induces current harmonics and mismatches between permanent magnet synchronous machine reference voltages and terminal voltages, which will degrade the sensorless drive system performance, especially at the low-speed range. In this paper, a radial basis function neural network (RBFNN)-based voltage compensator is proposed to suppress the current ripple. Without the requirement of any additional hardware or complex signal analysis procedure and processing algorithm, the RBFNN is self-tuned to directly generate the compensation voltage with the objective to minimize the current tracking error, so as to improve the active flux modeling accuracy and reduce position and speed estimation fluctuation.

Multiple object tracking (MOT) is an important task in computer vision, especially video analytics. Transformer-based methods are emerging approaches using both tracking and detection queries. However, motion modeling in existing transformer-based methods lacks effective association capability. Thus, this paper introduces a new METFormer model, a Motion Enhanced TransFormer-based tracker with a novel global-local motion context learning technique to mitigate the lack of motion information in existing transformer-based methods. The global-local motion context learning technique first centers on difference-guided global motion learning to obtain temporal information from adjacent frames. Based on global motion, we leverage context-aware local object motion modelling to study motion patterns and enhance the feature representation for individual objects. Experimental results on the benchmark MOT17 dataset show that our proposed method can surpass the state-of-the-art Trackformer [21] by 1.8% on IDF1 and 21.7% on ID Switches under public detection settings.

Early-stage fire and smoke detection through visual analysis is crucial for industrial safety and hazard prevention. However, detecting fire and smoke in factories using surveillance cameras poses challenges due to the small size of target objects. To address these challenges, we introduce a refined single-stage detector called FFS-YOLO (Factory Fire Smoke - YOLO). Our approach incorporates the Parameter-Free Attention Module (SimAM) and ResNet-SimMix module into the Backbone and Head of YOLOv7 to enhance key feature extraction. Additionally, we modify the model architecture by adding an extra prediction head to facilitate the fusion of features at multiple scales, specifically for small-scale object detection. Experimental results conducted on our fire and smoke dataset demonstrate the effectiveness of the FFS-YOLO model, achieving an average mAP, Precision, and Recall of 0.92, 0.91, and 0.90, respectively. The performance of the proposed model outperforms existing relevant competitors in the field. The findings of this research contribute significantly to the advancement of early fire detection and prevention in factory settings.

In this paper, the influence of full-pitch and shortpitch windings on the torque and power factor of permanentmagnet vernier (PMV) machines is studied.It is well known that the short-pitch (SP) winding has shorter end windings but with a lower winding factor.Thus, to achieve the largest torque, the full-pitch (FP) winding is preferred in PMV machines.However, it is found in this study that the SP winding, rather than the FP winding, is a better choice to achieve higher output torque of PMV machine at a high power factor level under the same PM consumption.The advantages of SP winding are even more obvious at higher electric loading and higher power factor levels.The theoretical analysis indicates that both torque and power factors are influenced by the winding factor, providing the possibility of better performance of the SP winding than the FP winding.This is further confirmed by the optimization results after performing the multi-objective optimization considering torque, power factor, and PM consumption.Finally, a prototype is manufactured and tested to validate the analysis.

This paper presents a high power-factor permanent magnet vernier machine (PMVM) employing hybrid concentrated-winding (CW). The hybrid winding, carrying both delta- and star-winding sets, allows for low harmonic tooth-wound coil and satisfied winding factor for PMVM with high gear ratio. Therefore, the torque density of the PMVM can be enhanced with reduced end-winding length and more compact structure. More importantly, the mutual inductances between the three phases and the coils that comprise each phase are reduced to a large extent in the hybrid winding, without deteriorating the output torque capability. As a result, the presented PMVM possesses a lower ratio of inductance to PM flux linkage due to the low-harmonic and low-coupling winding, and hence, improved power factor. Two benchmark PMVMs with identical gear ratio have been optimized for comparison. Further finite element results verify that the proposed CW PMVM presents superior performance in terms of end-winding length, torque density, power factor, and efficiency. It is revealed the proposed hybrid CW PMVM can improve the torque density to 23Nm/L from 21Nm/L and 14Nm/L, with the consideration of end-winding volume, and power factor to 0.86 from 0.72.

This paper investigates the two-slot pitch winding vernier machine by stator shifting technique (SST). It shows that the conventional two-slot pitch winding vernier machine can be generated by SST with a special stator shift angle. This paper contributes to exploring other shift angles and their influences on machine performances. Analysis results indicate that different shift angles have a great effect on the armature winding magnetomotive force (MMF) harmonic contents and the flux modulation effect. Consequently, the phase inductance, power factor, core losses, and torque capability will be affected. A single-layer 24-slot/10-pole vernier machine is investigated for verification, and the results show that with an appropriate shift angle, the power factor is improved from 0.6 to 0.7, the iron losses are reduced by 24.5%, and the efficiency is improved by 0.7%, with a slight torque drop of 2.9%.

Vehicle following is one of the fundamental functions of an autonomous driving system. Detection and recognition of tail light signal is important to prevent an autonomous vehicle from rear-end collisions or accidents. Although sensors like acoustic sonar or commercialized Advanced Driving Assistance System (ADAS) products such as mobileye could be used for rear-end collision warning, a cost-effective approach is expected. In this paper, we have developed a novel two-stage approach to detect vehicles and recognize brake lights from a single image in real-time. Unlike previous approaches where pair taillight has to be extracted explicitly, we use vehicle rear appearance image instead. On a large database, “Brake Lights Patterns” (BLP) are learned by a multi-layer perception neural network. Given an image, the vehicles can be classified as “brake” or “normal” using the deep classifier. The vehicle can be detected quickly and robustly by combining multi-layer lidar (IBEO Lux fusion system) and a camera. Road segmentation and a novel vanishing point region of interest (ROI) determination method are explored to further speed up the detection and improve the robustness. The experimental results conducted on some real on-road videos have shown the robustness and efficiency of the proposed approach.

This paper proposes a map free lane following solution based on low-cost 2D laser scanners for Autonomous Service Vehicle to fill the gap between future driverless car and the lane keeping assistant. The applications of autonomous service vehicle include feeder bus in a local residential area, shuttle bus in a park or playground, sprinkler car, sweeper car, and transporter in airport or container terminal. As autonomous service vehicle is running only in a limited area and its speed is slow compared to normal vehicles, we can further simplify the problem regardless of the issues of road infrastructure detection/communication and V2I maps which prevent the popularization of driverless car, and to propose a unique map free solution. The features of our approach include: 1) an innovative configuration for two 2D laser scanners to detect the lane with sharp curve; 2) a fast and accurate lane detection algorithm based on 2D laser's raw date directly; 3) a reliable and smooth path planning based on local lane fitting and prediction; and 4) a self-built unique drive-by-wire system for electronic car. We successfully tested our vehicle with autonomous driving in the testing field. The experiments show that the vehicle's trajectory matched the planned path accurately.

This paper discusses and proposes a method to minimize the effect of lateral motion of a humanoid robot's body that hampers localization performance of an adult-size humanoid robot. On a humanoid with a large sway motion, this effect is not negligible and a correction on the position estimation for perception based localization systems is necessary. First, the sway motion will be quantified and different perception configurations will be proposed to reduce the errors caused by this motion. Then, experimental results will be presented to demonstrate that the proposed localization system using a 2D laser scanner is more accurate and reliable when the sway motion is compensated.

It is the aim of this work to research how short-comings of a social robot due to its expressive limitations may be overcome by multimodal feedback. An experiment is proposed in which a robot that cannot produce facial expressions plays a game of rock, paper, scissors with people. A screen which is built-in the torso of the robot is used to compensate for these limitations in expressiveness and provide the participant with facial expressions during the game. To assess the impact of the screen on the user's rating of the robot and the interaction as such, there will be a control condition in which the screen will stay turned off. With this experiment we would like to show that sophisticated feedback setup can contribute to make a playful interaction between a human and a robot even more enjoyable.

In order to provide a user-friendly system with simple operation command to grasp different objects successfully, this paper describes a combined approach of real time remote vision-based teleoperation and autonomy for a human-aided robotic grasping. In the teleoperation process, motion tracking is carried out by Kinect in real time to detect the positions of the human shoulder, elbow and hand joints such that the robot can imitate the human. Hand gestures are recognized and used to activate autonomous grasping, which can save time and generate more natural grasping poses. In our system, the robot fulfills some special tasks such as picking up objects using easy commands with Kinect as object sensor. Experiment results show that it is effective and user-friendly.

It is believe that developing a robot without considering users feedback on appearance, functions and behavior will cause a share in the failure of research and evaluation of human robot interaction. This paper researched and solicited feedback from targeted users on their expectations and views of a robot, which will be used as a receptionist in a Singapore research centre. Based on design inputs from BMW Group DesignworksUSA, an experiment with 36 participants was conducted. The study covered areas such as perceptions of a robot, concerns of developing a robotic receptionist, expectations for a robotic receptionist, and preference in terms of overall appearance and functionalities of the intended robotic receptionist. The approach to the robotic development was based on a design methodology that took into consideration elements that would improve human robot interaction. The robotic receptionist was designed with a strong emphasis on the robot's appearance, behavior and functions that would impact the end users. The intent was to create a pleasurable experience for the end users during their encounter with the robot. This study finally led to conceptual designs based on response and selection from the participants. The robotic receptionist design had been selected and the first prototype, Olivia 2.0 has been developed for further human robot interaction studies.

This paper describe an implemented robot lion dance system, developed by a multi-disciplinary team of researchers over the past three years. We aim to use advance robotic technology to develop a mechatronic system that can perform life-size lion dancing with the traditional lion dance outfit. We intend to experiment the fusion of traditional art form and robotic technology so as to stimulate people's interest in the disappearing traditional art form and folk lords, and give new meanings to the new art form. The robot employs an event design of the robotic lion head, upper and lower body and methodology of conveying robotic lion dance into robotic lion motion. In this paper we will first describe the design concept and the architecture of the robot lion. Secondly, we will describe in detail the workings of the robotic lion dance choreography and motion control to mimic a life size lion dancing. Lastly, this paper will highlight three key challenges that were faced during the design of the robot to mimic the actual human lion dance.