Search

This paper presents the ACC system for autonomous trams, including three core processes: object detection, collision assessment and control. The system utilizes the YOLOX deep learning model for camera-based object detection and a clustering approach for LiDAR point clouds, integrating both outputs through a late fusion method for enhanced accuracy. The positions of objects and the tram are projected onto a data map to assess collision risk. This data map, which is a virtual 2D space configured with the track’s UTM coordinates and ROI lines, is used to determine whether objects are located within the ROI. The ACC system adjusts the tram’s speed to control acceleration or deceleration based on the relative distances of objects within the ROI. Experiments on a test track demonstrate effective collision avoidance and reliable ACC system operation under predefined scenarios.

To achieve Level 4 and above autonomous driving, a robust and stable autonomous driving system is essential to adapt to various environmental changes. This paper aims to perform vehicle pose estimation, a crucial element in forming autonomous driving systems, more universally and robustly. The prevalent method for vehicle pose estimation in autonomous driving systems relies on Real-Time Kinematic (RTK) sensor data, ensuring accurate location acquisition. However, due to the characteristics of RTK sensors, precise positioning is challenging or impossible in indoor spaces or areas with signal interference, leading to inaccurate pose estimation and hindering autonomous driving in such scenarios. This paper proposes a method to overcome these challenges by leveraging objects registered in a high-precision map. The proposed approach involves creating a semantic high-definition (HD) map with added objects, forming object-centric features, recognizing locations using these features, and accurately estimating the vehicle’s pose from the recognized location. This proposed method enhances the precision of vehicle pose estimation in environments where acquiring RTK sensor data is challenging, enabling more robust and stable autonomous driving. The paper demonstrates the proposed method’s effectiveness through simulation and real-world experiments, showcasing its capability for more precise pose estimation.

Deep learning based models on the edge devices have received considerable attention as a promising means to handle a variety of AI applications. However, deploying the deep learning models in the production environment with efficient inference on the edge devices is still a challenging task due to computation and memory constraints. This paper proposes a framework for the service robot named GuardBot powered by Jetson Xavier NX and presents a real-world case study of deploying the optimized face mask recognition application with real-time inference on the edge device. It assists the robot to detect whether people are wearing a mask to guard against COVID-19 and gives a polite voice reminder to wear the mask. Our framework contains dual-stage architecture based on convolutional neural networks with three main modules that employ (1) MTCNN for face detection, (2) our proposed CNN model and seven transfer learning based custom models which are Inception-v3, VGG16, denseNet121, resNet50, NASNetMobile, XceptionNet, MobileNet-v2 for face mask classification, (3) TensorRT for optimization of all the models to speedup inference on the Jetson Xavier NX. Our study carries out several analysis based on the models’ performance in terms of their frames per second, execution time and images per second. It also evaluates the accuracy, precision, recall & F1-score and makes the comparison of all models before and after optimization with a main focus on high throughput and low latency. Finally, the framework is deployed on a mobile robot to perform experiments in both outdoor and multi-floor indoor environments with patrolling and non-patrolling modes. Compared to other state-of-the-art models, our proposed CNN model for face mask recognition based on the classification obtains 94.5%, 95.9% and 94.28% accuracy on training, validation and testing datasets respectively which is better than MobileNet-v2, Xception and InceptionNet-v3 while it achieves highest throughput and lowest latency than all other models after optimization at different precision levels.

Pedestrian tracking is a challenging task in the area of visual object tracking research and it is a vital component of various vision-based applications such as surveillance systems, human-following robots, and autonomous vehicles. In this paper, we proposed a single pedestrian tracking (SPT) framework for identifying each instance of a person across all video frames through a tracking-by-detection paradigm that combines deep learning and metric learning-based approaches. The SPT framework comprises three main modules: detection, re-identification, and tracking. Our contribution is a significant improvement in the results by designing two compact metric learning-based models using Siamese architecture in the pedestrian re-identification module and combining one of the most robust re-identification models for data associated with the pedestrian detector in the tracking module. We carried out several analyses to evaluate the performance of our SPT framework for single pedestrian tracking in the videos. The results of the re-identification module validate that our two proposed re-identification models surpass existing state-of-the-art models with increased accuracies of 79.2% and 83.9% on the large dataset and 92% and 96% on the small dataset. Moreover, the proposed SPT tracker, along with six state-of-the-art (SOTA) tracking models, has been tested on various indoor and outdoor video sequences. A qualitative analysis considering six major environmental factors verifies the effectiveness of our SPT tracker under illumination changes, appearance variations due to pose changes, changes in target position, and partial occlusions. In addition, quantitative analysis based on experimental results also demonstrates that our proposed SPT tracker outperforms the GOTURN, CSRT, KCF, and SiamFC trackers with a success rate of 79.7% while beating the DiamSiamRPN, SiamFC, CSRT, GOTURN, and SiamMask trackers with an average of 18 tracking frames per second.

This paper presents a state observer for an elastic joint with nonlinear friction via the information from an acceleration sensor. In order to avoid discontinuities, the nonlinear friction of the motor, which includes static, coulomb, and viscous terms, is considered a smooth function. In addition, it uses an acceleration sensor to obtain the information about the link with high uncertainty. The proposed state observer guarantees that the estimation error for the position and velocity of the link connected via an elastic joint containing a nonlinear stiffness (elasticity) converges to zero. In addition, it is shown that the observer gain can be designed by LMI (linear matrix inequality) optimization. Finally, to verify the performance of the proposed observer, the method proposed in this paper is tested by experiments on a two-inertia system with an elastic shaft.

Autonomous driving and its real-world implementation have been among the most actively studied topics in the past few years. In recent years, this growth has been accelerated by the development of advanced deep learning-based data processing technologies. Moreover, large automakers manufacture vehicles that can achieve partially or fully autonomous driving for driving on real roads. However, self-driving cars are limited to some areas with multi-lane roads, such as highways, and self-driving cars that drive in urban areas or residential complexes are still in the development stage. Among autonomous vehicles for various purposes, this paper focused on the development of autonomous vehicles for garbage collection in residential areas. Since we set the target environment of the vehicle as a residential complex, there is a difference from the target environment of a general autonomous vehicle. Therefore, in this paper, we defined ODD, including vehicle length, speed, and driving conditions for the development vehicle to drive in a residential area. In addition, to recognize the vehicle’s surroundings and respond to various situations, it is equipped with various sensors and additional devices that can notify the outside of the vehicle’s state or operate it in an emergency. In addition, an autonomous driving system capable of object recognition, lane recognition, route planning, vehicle manipulation, and abnormal situation detection was configured to suit the vehicle hardware and driving environment configured in this way. Finally, by performing autonomous driving in the actual experimental section with the developed vehicle, it was confirmed that the function of autonomous driving in the residential area works appropriately. Moreover, we confirmed that this vehicle would support garbage collection works through the experiment of work efficiency.

Simultaneous localization and mapping (SLAM) is an important function for service robots to self-navigate modernized buildings. However, only a few existing applications allow them to automatically move between stories through elevator. Some approaches have accomplished with the aid of hardware; however, this study shows that computer vision can be a promising alternative for button localization. In this paper, we proposed a real-time multi-story SLAM system which overcomes the problem of detecting elevator buttons using a localization framework that combines tracking and detecting approaches. A two-stage deep neural network initially locates the original positions of the target buttons, and a part-based tracker follows the target buttons in real-time. A positive-negative classifier and deep learning neural network (particular for button shape detection) modify the tracker's output in every frame. To allow the robot to self-navigate, a 2D grid mapping approach was used for the localization and mapping. Then, when the robot navigates a floor, the A* algorithm generates the shortest path. In the experiment, two dynamic scenes (which include common elevator button localization challenges) were used to evaluate the efficiency of our approach, and compared it with other state-of-the-art methods. Our approach was also tested on a prototype robot system to assesses how well it can navigate a multi-story building. The results show that our method could overcome the common background challenges that occur inside an elevator, and in doing so, it enables the mobile robot to autonomously navigate a multi-story building.

Object detection is an important task for rapidly localizing target objects using high-resolution satellite imagery (HRSI). Although deep learning has been shown an efficient means of detection, object detection in HRSI remains problematic due to variations in object scale and size. In this article, we present a novel deep neural network (DNN) that combines double-shot neural network with misplaced localization strategy that adapts to object detection tasks in satellite images. This novel architecture optimizes the localization of small and narrow rectangular objects, which frequently appear in HRSI images, without accuracy loss on other size and width/height ratio objects. This method outperforms other state-of-art methods. We evaluated our proposed method on the NWPU VHR-10 public dataset and a new benchmark dataset (seven classes of small and narrow rectangular objects, SNRO-7). The NWPU VHR-10 dataset built a dataset for multiclass object detection; however, most labels are assigned in normal size and width/height ratios. SNRO-7 focuses on multiscale and multisize object detection and includes many small-size and narrow rectangular objects. We also evaluated the accuracy difference on DNN training and testing between gray scale and RGB datasets. The results of the experiment on object detection reveal that the mean average precision (MaP) of our method is 82.6% in NWPU VHR-10 and 79.3% in SNRO-7, which exceeds the MaPs of other state-of-the-art object detection neural networks. The model trained with the RGB dataset can achieve similar accuracy (around 79.0% MIoU) testing in both RGB and gray scale datasets. When training the model by mixing RGB and gray scale datasets in different ratios, the accuracy in the RGB channel significantly decreases with increasing gray scale images, but this does not influence the accuracy in the gray scale dataset.

In this paper, we propose an adaptive backstepping controller to control the exact position and orientation of a remotely operated underwater vehicle with parametric model uncertainty. To further improve the angular velocity control precision of each thruster, a phase locked loop (PLL) controller has been added to the backstepping controller. A comparison of two backstepping controllers with and without the PLL control loop has been performed using simulations and experiments. The test results showed that the tracking performance could be improved by using the PLL control loop in the proposed adaptive backstepping controller.

3D visual recognition is a prerequisite for most autonomous robotic systems operating in the real world. It empowers robots to perform a variety of tasks, such as tracking, understanding the environment, and human–robot interaction. Autonomous robots equipped with 3D recognition capability can better perform their social roles through supportive task assistance in professional jobs and effective domestic services. For active assistance, social robots must recognize their surroundings, including objects and places to perform the task more efficiently. This article first highlights the value-centric role of social robots in society by presenting recently developed robots and describes their main features. Instigated by the recognition capability of social robots, we present the analysis of data representation methods based on sensor modalities for 3D object and place recognition using deep learning models. In this direction, we delineate the research gaps that need to be addressed, summarize 3D recognition datasets, and present performance comparisons. Finally, a discussion of future research directions concludes the article. This survey is intended to show how recent developments in 3D visual recognition based on sensor modalities using deep-learning-based approaches can lay the groundwork to inspire further research and serves as a guide to those who are interested in vision-based robotics applications.

Knowledge representation in autonomous robots with social roles has steadily gained importance through their supportive task assistance in domestic, hospital, and industrial activities. For active assistance, these robots must process semantic knowledge to perform the task more efficiently. In this context, ontology-based knowledge representation and reasoning (KR & R) techniques appear as a powerful tool and provide sophisticated domain knowledge for processing complex robotic tasks in a real-world environment. In this article, we surveyed ontology-based semantic representation unified into the current state of robotic knowledge base systems, with our aim being three-fold: (i) to present the recent developments in ontology-based knowledge representation systems that have led to the effective solutions of real-world robotic applications; (ii) to review the selected knowledge-based systems in seven dimensions: application, idea, development tools, architecture, ontology scope, reasoning scope, and limitations; (iii) to pin-down lessons learned from the review of existing knowledge-based systems for designing better solutions and delineating research limitations that might be addressed in future studies. This survey article concludes with a discussion of future research challenges that can serve as a guide to those who are interested in working on the ontology-based semantic knowledge representation systems for autonomous robots.

Humans have an innate ability of environment modeling, perception, and planning while simultaneously performing tasks. However, it is still a challenging problem in the study of robotic cognition. We address this issue by proposing a neuro-inspired cognitive navigation framework, which is composed of three major components: semantic modeling framework (SMF), semantic information processing (SIP) module, and semantic autonomous navigation (SAN) module to enable the robot to perform cognitive tasks. The SMF creates an environment database using Triplet Ontological Semantic Model (TOSM) and builds semantic models of the environment. The environment maps from these semantic models are generated in an on-demand database and downloaded in SIP and SAN modules when required to by the robot. The SIP module contains active environment perception components for recognition and localization. It also feeds relevant perception information to behavior planner for safely performing the task. The SAN module uses a behavior planner that is connected with a knowledge base and behavior database for querying during action planning and execution. The main contributions of our work are the development of the TOSM, integration of SMF, SIP, and SAN modules in one single framework, and interaction between these components based on the findings of cognitive science. We deploy our cognitive navigation framework on a mobile robot platform, considering implicit and explicit constraints for autonomous robot navigation in a real-world environment. The robotic experiments demonstrate the validity of our proposed framework.

This paper proposes a method for mobile robot localization in a partially unknown indoor environment. The method fuses two types of range measurements: the range from the robot to the beacons measured by ultrasonic sensors and the range from the robot to the walls surrounding the robot measured by a laser range finder (LRF). For the fusion, the unscented Kalman filter (UKF) is utilized. Because finding the Jacobian matrix is not feasible for range measurement using an LRF, UKF has an advantage in this situation over the extended KF. The locations of the beacons and range data from the beacons are available, whereas the correspondence of the range data to the beacon is not given. Therefore, the proposed method also deals with the problem of data association to determine which beacon corresponds to the given range data. The proposed approach is evaluated using different sets of design parameter values and is compared with the method that uses only an LRF or ultrasonic beacons. Comparative analysis shows that even though ultrasonic beacons are sparsely populated, have a large error and have a slow update rate, they improve the localization performance when fused with the LRF measurement. In addition, proper adjustment of the UKF design parameters is crucial for full utilization of the UKF approach for sensor fusion. This study contributes to the derivation of a UKF-based design methodology to fuse two exteroceptive measurements that are complementary to each other in localization.

Person-following technology is an important robot service. The major trend of person-following is to utilize computer vision technology to localize the target person, due to the wide view and rich information that is obtained from the real world through a camera. However, most existing approaches employ the detecting-by-tracking strategy, which suffers from low speed, accompanied with more complicated detecting models and unstable region of interest (ROI) outputs in unexpressed situations. In this paper, we propose a novel classification-lock strategy to localize the target person, which incorporates the visual tracking technology with object detection technology, to adapt the localization model to different environments online, and to keep a high frame-per-second (FPS) on the mobile platform. This person-following approach consists of three key parts. In the first step, a pairwise cluster tracker is employed to localize the person. A positive and negative classifier is then utilized to verify the tracker’s result and to update the tracking model. In addition, a detector pre-trained by a CPU-optimized convolutional neural network is used to further improve the result of tracking. In the experiment, our approach is compared with other state-of-art approaches by a Vojir tracking dataset, with three sequences in the items of human to prove the quality of person localization. Moreover, the common challenges during the following task are evaluated by several image sequences in a static scene, and a dynamic scene is used to evaluate the improvement from the classification-lock strategy. Finally, our approach is deployed on a mobile robot to test its performance on the function of the person-following. Compared with other state-of-art methods, our approach achieves the highest score (0.91 recall rate). In the static and dynamic scene, the output of the ROI based on the classification-lock strategy is significantly better than that without it. Our approach also succeeds in a long-term following task in an indoor multi-floor scenario.

In order to prove the speed regulation for permanent magnet motor systems, a control method is proposed to reject the disturbance by cogging torque, which arises from the interaction of the rotor magnets with the steel teeth on the stator. Toward this aim, we present a state observer with an internal model which has harmonics of sinusoids, and it follows from the proposed observer that the information of the cogging disturbance is obtained. It shows that the disturbance is efficiently compensated by a suitable additional control input based on the estimated information. Simulation and experimental results with an actual permanent magnet motor verify the effectiveness of the proposed observer-based algorithm.

In this paper, we consider the state estimation problem for flexible joint manipulators that involve nonlinear characteristics in their stiffness. The two key ideas of our design are that (a) an accelerometer is used in order that the estimation error dynamics do not depend on nonlinearities at the link part of the manipulators and (b) the model of the nonlinear stiffness is indeed a Lipschitz function. Based on the measured acceleration, we propose a nonlinear observer under the Lipschitz condition of the nonlinear stiffness. In addition, in order to effectively compensate for the estimation error, the gain of the proposed observer is chosen from the ARE (algebraic Riccati equations) which depend on the Lipschitz constant. Comparative experimental results verify the effectiveness of the proposed method.