Search

In this chapter, we presented a novel photometric quasi-invariant SIFT descriptor, PHI-SIFT, which is insensitive to photometric variations in addition to geometric invariance. The dichromatic reflection model is used for extracting photometric quasi-invariant features and the similar approach to the SIFT is used for obtaining geometric invariance. For performance evaluation, we compared this descriptor with the SIFT and the CSIFT descriptors. Experiments show that our method gives similar performance or outperforms them with respect to stability and distinctiveness. This method may be applicable to image retrieval, object detection and recognition.

This paper describes a novel object tracking method which combines a conventional global nearest-neighbor (GNN) method with an enhanced data association approach using 3D data features for an unmanned ground vehicle. If an object crosses in front of an observation system which has a measurement sensor equipped on a vehicle, a conventional distance-based object tracking mechanism encounters a data association problem owing to high uncertainties and a leak of information: correspondence between observations and tracks. To overcome these correspondence problems, we use the object’s 3D information such as statistical and geometrical information. In this paper, we describe a clustering and segmentation processes in conjunction with a distance-based nearest-neighbor approach, RBNN-2, and applied GNN with enhanced data association using both a kinematic and geometric models derived from 3D range data. To evaluate our method using a 3D LADAR sensor, a Velodyne HDL-32e, installed on the vehicle, we conducted an experiment involving two crossing humans and a ground vehicle to identify different human trajectories for a leader-following system.

This paper investigates techniques on improving navigation accuracy using multiple sensors mounted on a mobile platform and exploiting the inherent characteristic of a ground vehicle that does not move along the cross-track and off the ground, often termed nonholonomic constraints (NHC) for car-like vehicles that assume no slip or skid. The forward velocity of the vehicle is obtained using a wheel encoder. The 3D velocity vector becomes observable during the normal moving state of the vehicle by using NHC, which produces one virtual sensor. Another virtual sensor is the zero-velocity update (ZVU) condition of the vehicle; when the condition is true, the 3D velocity vector (which is zero) becomes observable. These observables were employed in an extended Kalman filter (EKF) update to limit the growth of inertial navigation system error. We designed an EKF for data fusion of inertial measurement units, global positioning systems (GPS), and motion constraints (i.e., NHC and ZVU). We analyzed the effects of utilizing these constraints on improving navigation accuracy in stationary and dynamic cases. Our proposed navigation suite provides reliable accuracy for unmanned ground vehicle applications in a GPS-denied environment (e.g., forest canopy and urban canyon).

Pedestrian navigation systems (PNS) using foot-mounted MEMS inertial sensors use zero-velocity updates (ZUPTs) to reduce drift in navigation solutions and estimate inertial sensor errors. However, it is well known that ZUPTs cannot reduce all errors, especially as heading error is not observable. Hence, the position estimates tend to drift and even cyclic ZUPTs are applied in updated steps of the Extended Kalman Filter (EKF). This urges the use of other motion constraints for pedestrian gait and any other valuable heading reduction information that is available. In this paper, we exploit two more motion constraints scenarios of pedestrian gait: (1) walking along straight paths; (2) standing still for a long time. It is observed that these motion constraints (called “virtual sensor”), though considerably reducing drift in PNS, still need an absolute heading reference. One common absolute heading estimation sensor is the magnetometer, which senses the Earth’s magnetic field and, hence, the true heading angle can be calculated. However, magnetometers are susceptible to magnetic distortions, especially in indoor environments. In this work, an algorithm, called magnetic anomaly detection (MAD) and compensation is designed by incorporating only healthy magnetometer data in the EKF updating step, to reduce drift in zero-velocity updated INS. Experiments are conducted in GPS-denied and magnetically distorted environments to validate the proposed algorithms.

This paper provides algorithms to estimate absolute navigation information, e.g., absolute attitude and position, by using low power, weight and volume Microelectromechanical Systems-type (MEMS) sensors that are suitable for micro planetary rovers. Planetary rovers appear to be easily navigable robots due to their extreme slow speed and rotation but, unfortunately, the sensor suites available for terrestrial robots are not always available for planetary rover navigation. This makes them difficult to navigate in a completely unexplored, harsh and complex environment. Whereas the relative attitude and position can be tracked in a similar way as for ground robots, absolute navigation information, unlike in terrestrial applications, is difficult to obtain for a remote celestial body, such as Mars or the Moon. In this paper, an algorithm called the EASI algorithm (Estimation of Attitude using Sun sensor and Inclinometer) is presented to estimate the absolute attitude using a MEMS-type sun sensor and inclinometer, only. Moreover, the output of the EASI algorithm is fused with MEMS gyros to produce more accurate and reliable attitude estimates. An absolute position estimation algorithm has also been presented based on these on-board sensors. Experimental results demonstrate the viability of the proposed algorithms and the sensor suite for low-cost and low-weight micro planetary rovers.

A new method of obtaining geometric features of three-dimensional range data for human and tree recognition in an off-road environment is described. The learning algorithm AdaBoost is used to select a set of discriminative features from a very large set of potential geometric features. The proposed geometric feature can be considered as a generalisation of the geometric feature used in previous studies. The experimental results for human and tree recognition show that the proposed method outperforms the other methods.

Autonomous navigation is one of primitive functionalities which service robots should have; nevertheless, the navigation problem for a service robot still has many difficulties because the real environment where service robots should works is so complex and dynamical. This paper proposes a framework of intelligent navigation for service robots based on a semantic map of the smart environment. In the smart environment, the robot can receive his position information from location sensors by sensor networks, and it can eliminate the accumulated localization errors. So the robot can accomplish confidential localization in the dynamic and complex environment. With the topological information in the semantic map, path planning problem can be simple even in the wide and complex spaces, and topological path from semantic map can be divided into several sub goals. Robust navigation can be accomplished by moving towards these sub goals with reactive navigation algorithm which has robust characteristics in the dynamic environments. Our approaches have ascertained the good performance on the localization and the navigation, and the feasibility of these methods could be confirmed with the result of experiments in the real environment.

Position and attitude estimation of helmet-mounted imaging devices e.g. camera/lidar is difficult in unstructured indoor environment due to lack of conventional localization systems, e.g. RF, Ultrasonic, UWB and Wi-Fi signals, usually available in modern office-like buildings. In this work, we use single MEMS IMU fitted on foot, which when combined with zero-velocity updates (ZUPT) in Extended Kalman filter estimation framework at every foot step, provides very accurate position estimates, regardless of the user and environment. We also present a novel method to reduce the error drift in ZUPT-only position estimates by employing the pitching motion of the foot during the swing phase. Another small IMU is attached on helmet, which can provide attitude information of imaging device at most at its own. However, for proper mapping applications, both position and attitude information of the imaging device is required at high rate, which is difficult to obtain from helmet IMU alone. To get complete pose information of the imaging device, we make use of the so called, ‘Transfer Alignment’ techniques, borrowed from avionics community. Experimental results show that poses of the imaging device is obtained with sufficient accuracy for mapping application without any extra sensors network aiding.

This paper aims at developing algorithmic approach to reduce the error drift in navigation information using only single motion sensor i.e. inertial measurement unit (IMU) in an unstructured environment. The target application is pedestrian navigation system (PNS), but these algorithms are general for other applications too involving autonomous systems. A MEMS IMU is mounted on the foot of the agent in order to get benefit of known walking patterns of the wearer. Algorithms are derived based on motion constraints of human walking and applied to reduce the inherent error drift in inertial navigation systems (INS). Experiments are conducted in indoor/outdoor unstructured environment to validate this algorithmic approach and shown that without using any other sensors and any other pre-conditions on the environment; we can reduce navigation errors significantly only by taking into account the motion constraints of human walking.

This paper describes the occluded target recognition and tracking method for a leader-following system by fusing 3D range and image data acquired from 3D light detection and ranging (LIDAR) and a color camera installed on an autonomous vehicle in forest environment. During 3D data processing, the distance-based clustering method has an instinctive problem in close encounters. In the tracking phase, we divide an object tracking process into three phases based on occlusion scenario; before an occlusion (BO) phase, a partially or fully occlusion phase and after an occlusion (AO) phase. To improve the data association performance, we use camera's rich information to find correspondence among objects during above mentioned three phases of occlusion. In this paper, we solve a correspondence problem using the color features of human objects with the sum of squared distance (SSD) and the normalized cross correlation (NCC). The features are integrated with derived windows from Harris corner. The experimental results for a leader following on an autonomous vehicle are shown with LIDAR and camera for improving a data association problem in a multiple object tracking system.

This chapter explores using service-oriented computing to reengineer non-secure legacy software applications to create new secure target applications. Two objectives of this chapter are: (1) to analyze the architectural changes required in order to adopt new web technologies and cope with resultant vulnerabilities in source code; and (2) to measure the level of effort required to modernize software by adopting new web technologies and adding security countermeasures. To meet these objectives, a model-driven Scrum for Service-Oriented Software Reengineering (mScrum4SOSR) methodology was chosen and applied to a reengineering project. Scrum is employed to manage the reengineering project, as well as to measure implementation effort related to the modernization process. Further, a re-documentation technique called 5W1H Re-Doc is used to re-document the non-secure software application at a high level of abstraction in order to help project participants comprehend what is needed to identify candidate services for service-oriented reengineering. Case studies with and without security features are created for different types of applications - a desktop graphical user interface, a web application, a web services application, a restful web services application, and an enterprise service bus application.

The Joint Architecture for Unmanned Systems (JAUS) is a communication standard that allows for interoperability between Unmanned Vehicles (UVs). Current research indicates that JAUS-compliant systems do not meet real-time performance guidelines necessary for internal systems in UVs. However, there is a lack of quantitative data illustrating the performance shortcomings of JAUS or clear explanations on what causes these performance issues or comparisons with existing internal communication systems. In this research, we first develop a basic C++ implementation of JAUS and evaluate its performance with quantitative data and compare the results with published performance data of Controller Area Network (CAN) to determine the feasibility of the JAUS standard. Our results indicate that the main reason of JAUS’s poor performance lies in the latency inherent in the hierarchical structure of JAUS and the overhead of User Datagram Protocol (UDP) messages, which has been used with JAUS and is slower than the high-speed CAN. Additionally, UDP has no scheduling mechanism, which makes it virtually impossible to guarantee messages meeting their deadlines. Considering the slow and nondeterministic JAUS communication from subsystems to components, which is JAUS Level 3 compliance, we then propose a solution by bringing Ethernet for Control Automation Technology (EtherCAT) to add speed, deterministic feature, and security. The JAUS-EtherCAT mapping, which we called a JEBridge, is implemented into nodes and components. Both quantitative and qualitative results are provided to show that JEBridge and JAUS Level 3 compliance can bring not only interoperability but also reasonable performance to UVs.

This paper describes a real-time object detection and tracking method which is improved to occluded measurements. 3D point cloud data describes surrounding objects on one measurement frame. If two objects cross in front of an observing system, equipped 3D LIDAR on a vehicle, the object recognition mechanism has inherent problems. In this paper, we define that the system process consists of a detection part and a tracking part. The first problem occurs on the clustering process in the detection part. Two objects are liked as one object by a neighbor point approach. In our case, RBNN (Radial Bounded Nearest Neighbor) clustering method works with two parameters, radial bounded range and point number. The second problem happens on the tracking process. The previous detection process hands on wrong the object number. The tracking system also shows the ambiguity results with the fault tracking. It is frequently happened on occluded measurements. We proposed the enhanced object detection and tracking method using an object class and relative position. In this paper, we propose the system which has an enhanced tracking mechanism for approaching successful object tracking which prevents wrong detection and tracking. In experiment, we implement two humans and an autonomous ground vehicle to achieve a leader-following system.

This paper deals with estimating the attitude of a walking quadruped robot in the presence of engine noise. Contrary to wheeled robots, legged robots are inherently supposed to work in uneven environment. Therefore estimating the attitude of walking legged robot becomes the utmost important factor in designing such machines in order to control them. We have used IMU as the prime sensor for stabilized attitude estimation. In fact accelerometer is used for roll/pitch observations in stationary/quasi-stationary state. But this data is not suitable unless pre-processing. This involves the pre-filtering and error factor determination for raw IMU data. Engine vibration effects are removed by low-pass filter and error-factor are calculated by least square fit to raw IMU data. Finally results of applying these simple techniques show the stability of attitude estimation on quadruped walking robot in real experiment.

This paper describes natural terrain detection and SLAM algorithms using a LIDAR sensor for an unmanned ground vehicle. We describe how features are detected from natural terrain, and then localize the vehicle’s position and compose a map with the detected features. The LIDAR is equipped on the experimental vehicle to scan natural terrain. The scan data include many kinds of natural disturbances on uneven terrain: a banded tree, a branch of a tree, non-uniform size of bushes, and undefined or unexpected objects. We apply a RANSAC (RANdom SAmple Consensus) algorithm to discriminate ground point cloud data and object point cloud data, and then separate bush points and tree points by a combination of two algorithms, GMM (Gaussian Mixture Model) and EM (Expectation Maximization). The GMM and EM algorithms are for extracting features and classifying groups, respectively. We propose a double FCM (Fuzzy C-mean clustering) algorithm to robustly estimate the number of trees and their positions. The Extended Kalman Filter applied to simultaneous localization and mapping (EKF-SLAM) is composed of the extracted tree features. The mahalanobis distance is applied to retain consistency for feature correspondence, which is used data association. Finally, we show the results obtained from implementation of the approach on a tree-covered mountain.

This paper describes an enhanced fusion method for an Inertial Navigation System (INS) based on a 3-axis accelerometer sensor, a 3-axis gyroscope sensor and a laser scanner. In GPS-denied environments, indoor or dense forests, a pure INS odometry is available for estimating the trajectory of a human or robot. However it has a critical implementation problem: a drift error of velocity, position and heading angles. Commonly the problem can be solved by fusing visual landmarks, a magnetometer or radio beacons. These methods are not robust in diverse environments: darkness, fog or sunlight, an unstable magnetic field and an environmental obstacle. We propose to overcome the drift problem using an Iterative Closest Point (ICP) scan matching algorithm with a laser scanner. This system consists of three parts. The first is the INS. It estimates attitude, velocity, position based on a 6-axis Inertial Measurement Unit (IMU) with both 'Heuristic Reduction of Gyro Drift' (HRGD) and 'Heuristic Reduction of Velocity Drift' (HRVD) methods. A frame-to-frame ICP matching algorithm for estimating position and attitude by laser scan data is the second. The third is an extended kalman filter method for multi-sensor data fusing: INS and Laser Range Finder (LRF). The proposed method is simple and robust in diverse environments, so we could reduce the drift error efficiently. We confirm the result comparing an odometry of the experimental result with ICP and LRF aided-INS in a long corridor.

In this paper, we proposed multiple object detection and classification methods using a multi-channel light detection and ranging (LIDAR) device for an unmanned ground vehicle (UGV) in dense forest. The natural terrain environment is described by trees and bushes which are representative wild objects. To describe them on a navigation map, we must detect and classify objects with a single LIDAR sensor rather than multiple sensors on an UGV. It has the advantage of price, communication, and synchronization. If scanning object data are similar with real data, it is better recognizable; a large amount of 3D point cloud data can resemble object in appearance. There is a trade-off relationship between point cloud data and acquisition time. Getting a number of items of point cloud data gradually increases with measurement time and computation time. Our approach for achieving multiple object detection and classification consists of three steps that start from the raw 3D point cloud data. The 3D point cloud data are composed of one frame by one scan. First, the point cloud data are divided by two groups, ground points and non-ground points. Second, the non-ground points are separated by vegetation points and object points. Finally, the object points are arranged as a set of feasible objects with a clustering method. In our approach, random sample consensus (RANSAC), support vector machine (SVM) with principle component analysis (PCA) and fuzzy c-mean (FCM) algorithms belong to each step to derive multiple objects detection and classification. In a local statistic approach, the density of a point cloud in the processing area influences the system performance. Our primary research goal is to achieve improved performance with a data set of one frame per scan for an UGV. Thus, we evaluate the proposed object detection and classification algorithm with a single scan frame which is acquired from three different LIDAR devices. The device has different scan channels. The 1-channel and 4-channel LIDAR devices are commercial devices, LMS-111 and LD-MRS. The 8-channel device, KIDAR-B25, is developed from our previous research with the collaboration of a Korean company. We verify our proposed method through three results: the object detection rate, the point classification rate and the computing time. The results of the method are analyzed with hand-labeled data.

At last year, we have been developing 3D scanning LIDAR designated as KIDAR-B25, which features the 3D scanning structure based on an optically and mechanically coupled instrument. In contrast with previous scanning LIDARs, vertical scanning is realized using two stepping motors synchronized with movement and moves in a spiral. From the results of outdoor experiments conducted last year to evaluate and measure the LIDAR performance and stability, we identified some limitations and problems that should be resolved. In the first instance, the samples per second are inefficient for use in detection, object clustering, and classification. In addition, the accuracy and precision of distance at every point is seriously affected by the reflectance and distance of the target. Therefore, we have focused on improving the 3D LIDAR range finding performance, speed of measurement, and stability regardless of environmental variation. Toward the realization of these goals, in this paper, we deal with two improvements compared with previous 3D LIDAR.

The purpose of this research is to develop a new 3D LIDAR sensor, named KIDAR-B25, for measuring 3D image information with high range accuracy, high speed and compact size. To measure a distance to the target object, we developed a range measurement unit, which is implemented by the direct Time-Of-Flight (TOF) method using TDC chip, a pulsed laser transmitter as an illumination source (pulse width: 10 ns, wavelength: 905 nm, repetition rate: 30kHz, peak power: 20W), and an Si APD receiver, which has high sensitivity and wide bandwidth. Also, we devised a horizontal and vertical scanning mechanism, climbing in a spiral and coupled with the laser optical path. Besides, control electronics such as the motor controller, the signal processing unit, the power distributor and so on, are developed and integrated in a compact assembly. The key point of the 3D LIDAR design proposed in this paper is to use the compact scanning mechanism, which is coupled with optical module horizontally and vertically. This KIDAR-B25 has the same beam propagation axis for emitting pulse laser and receiving reflected one with no optical interference each other. The scanning performance of the KIDAR-B25 has proven with the stable operation up to 20Hz (vertical), 40Hz (horizontal) and the time is about 1.7s to reach the maximum speed. The range of vertical plane can be available up to ±10 degree FOV (Field Of View) with a 0.25 degree angular resolution. The whole horizontal plane (360 degree) can be also available with 0.125 degree angular resolution. Since the KIDAR-B25 sensor has been planned and developed to be used in mobile robots for navigation, we conducted an outdoor test for evaluating its performance. The experimental results show that the captured 3D imaging data can be usefully applicable to the navigation of the robot for detecting and avoiding the moving objects with real time.

This paper describes an object detection and classification method for an Unmanned Ground Vehicle (UGV) using a range sensor and an image sensor. The range sensor and the image sensor are a 3D Light Detection And Ranging (LIDAR) sensor and a monocular camera, respectively. For safe driving of the UGV, pedestrians and cars should be detected on their moving routes of the vehicle. An object detection and classification techniques based on only a camera has an inherent problem. On the view point of detection with a camera, a certain algorithm should extract features and compare them with full input image data. The input image has a lot of information as object and environment. It is hard to make a decision of the classification. The image should have only one reliable object information to solve the problem. In this paper, we introduce a developed 3D LIDAR sensor and apply a fusion method both 3D LIDAR data and camera data. We describe a 3D LIDAR sensor which is developed by LG Innotek Consortium in Korea, named KIDAR-B25. The 3D LIDAR sensor detects objects, determines the object's Region of Interest (ROI) based on 3D information and sends it into a camera region for classification. In the 3D LIDAR domain, we recognize breakpoints using Kalman filter and then make a cluster using a line segment method to determine an object's ROI. In the image domain, we extract the object's feature data from the ROI region using a Haar-like feature method. Finally it is classified as a pedestrian or car using a trained database with an Adaboost algorithm. To verify our system, we make an experiment on the performance of our system which is mounted on a ground vehicle, through field tests in an urban area.

We propose an alternative approach for time interpolation of (Time to Digital Convertor)TDC in which a saw-tooth wave is generated instead of using sinusoidal wave in [10]. The proposed method avoids floating number operation for runtime as well as lowers the temperature variation caused by analog passive components. Since the differential portion of two projected signals is amplified, the wider range of time interval can be measured while preserving a few picoseconds resolution. Compared with the conventional analog interpolation, the proposed method using saw-tooth wave overcomes the drawbacks of conventional analog interpolation TDC by reducing the dependency on passive components which are sensitive to environment and operational conditions. It provides high repeatability and linearity property within a wide range of operation with high resolution of measurement. With the aid of proposed interpolation circuit, a practical TDC and its embedded measurement system are expected to be implemented. The experiment results demonstrate its feasibility and effectiveness for practical usage with a simple and low cost implementation.

The purpose of this paper is to propose a service-oriented reverse reengineering that consists of re-documentation and design recovery. To understand the legacy system, we need to re-document the legacy system in a highlevel abstraction by using UML visual models through an initial reverse engineering. For this re-documentation, we propose a new re-documentation technique called 5W1H (Why, Who, When, Where, What, and How) model-driven re-documentation. Based upon the re-documented software visual model, we attempt to identify the scope of locating the candidate services from a legacy system. To show the feasibility of our approach, two applications using different design methods and architectures are re-documented into visual models. After recovering design methods and architectural patterns from the visual models, which have been injected into source code implicitly or explicitly to bring clear separation of concerns, we locate the scope of candidate services from the legacy system. We can reduce the effort of identifying candidate services by using the visualized model since we can discover highly possible locations for reusable business logic effectively. The service-oriented reverse engineering using 5W1H re-documentation technique will help a service producer to easily discover the candidate services from a legacy system through software re-documentation.

The purpose of this paper is to demonstrate how a Virtual Enterprise Integration (VEI) project using Service-Oriented Architecture (SOA) and Enterprise Service Bus (ESB) can be effectively conducted by a virtual team of service brokers. Currently, VEI is accomplished through SOA and ESB using web services and business process engines that execute WSDL and WS-BPEL. To reengineer a Virtual Enterprise (VE) based on one or more legacy components, referred to as a legacy VE, the abstract and concrete parts of the relevant business processes of the VE need to be reverse engineered to a high level of abstraction. To develop new business processes, business process requirements need to be forward engineered into business processes in BPEL. However, service brokers need guidelines for comprehending the operations of the legacy VEs or for understanding the business process requirement. In order to provide clear communication of this information, we propose a UML model driven Business Process Development Methodology (BPDM) called mBPDM. We demonstrate its applicability and capabilities by applying it to two case studies: a loan application process system which involves reverse engineering and Washington State Patrol's Drug Recognition Evaluation system which involves forward engineering. Based upon the results of reverse and forward engineering of two virtual enterprise cases, the guidelines, which use UML as a blueprint with multi-architectural views, help service brokers understand the underlying process architecture and organization of a virtual enterprise that has been built using the SOA concept and the contemporary ESB.

Due to its restricted usage and accuracy in measuring extremely short interval of time, TDC (time to digital converter) embedded measurement devices are relatively expansive. Even if new applications are being developed in daily life as well as in industry, the cost condition of TDC based measurement systems needs to be improved. To reduce the deployment cost on a commercial scale, analogue interpolation method is preferred to digital interpolation method for most of TDCs. However, the analogue interpolation method does not provide robustness to temperature variation and non-uniformity of parts, which results in difficulty in maintaining the expected repeatability. In this paper, an auto-calibration method is proposed for robust operation and precise measurement of TDC under the condition of temperature variation and lack of parts uniformity.

For many years, various local descriptors that are insensitive to geometric changes such as viewpoint, rotation, and scale changes, have been attracting attention due to their promising performance. However, most existing local descriptors including the SIFT (Scale Invariant Feature Transform) are based on luminance information rather than color information thereby resulting in instability to photometric variations such as shadows, highlights, and illumination changes. In this paper, we propose a novel local descriptor, π-SIFT, that are invariant to both geometric and photometric variations. In order to achieve photometric invariance, we adopt photometric quasi-invariant features based on the dichromatic reflection model. The performance of the proposed descriptor is evaluated with SIFT.

Object pose estimation from stereo images with unknown correspondence is a thoroughly studied problem in the computer vision and robot engineering literatures. Especially, it is important to detect the desirable corresponding points from images for object pose estimation. For this, many approaches have been proposed. Among them, the local feature descriptor, which describe the feature points that are robust to image deformations in an object or image, is one of the most promising approaches that has been applied to the stable feature detection problem successfully. Although any descriptors including the SIFT represent superior performance, these are based on luminance information rather than color information thereby resulting in instability to photometric variations such as shadows, highlights, and illumination changes. Therefore, we propose a novel method which extracts the interest points that are insensitive to both geometric and photometric variations in order to estimate more accurate and desirable object pose. In this method, we use photometric quasi-invariant features based on the dichromatic reflection model in order to achieve photometric invariance, and the SIFT is used for geometric invariance as well. The performance of the proposed method is evaluated with other local descriptors. Experimental results show that our method gives similar performance or outperforms them with respect to various imaging conditions. Finally, we estimate object pose by using the features extracted via the proposed method.

A prototype smart home environment for service robots has been constructed in the research building of Korea Institute of Industrial Technology (KITECH) to demonstrate the feasibility of a robot-assisted future home environment. An inexpensive service robot outfitted with a camera, a Radio Frequency Identification (RFID) reader, and a communication module is installed in the building as a future service robot system which is cheap but robust. In our robotic system, the RFID reader gets the rough location data and object information, and then the robot performs the object recognition scheme to get the exact position of the object in order to grasp it. Due to its concise and unambiguous description of the complex media contents, we adopted MPEG-7 visual descriptors for our object recognition system. In this paper, we propose a fast object recognition system for our smart home environment project, on the basis of MPEG-7 visual descriptors including color and texture. Experimental results show that our proposed system works well with good performance in terms of speed and recognition rate. This object recognition system will be incorporated into our mobile robotics platform and a shape descriptor is to be included in the next version of our object recognition system.

Over the past few years, many research groups have attempted to build smart environments. Recently, several groups have been actively conducting research into the construction of smart home environment for service robots to be used as assistants. A home service robot must have are localization, navigation, object recognition, and object handling functionalities. These operations are usually performed by a service robot by itself and the robot can perform the given tasks in a limited environment. In addition, they have shown only limited capabilities in a natural environment. We initiated a smart home environment project, RoboMaidHome, for light-weight service robots to provide reliable services by interacting with the environment through wireless sensor networks. This environment consists of the following three main components: (i) smart objects and smart appliances with sensor network functionality; (ii) a home server that connects smart devices as well as maintains information for reliable services; and (iii) service robots that perform tasks in collaboration with the environment. In this paper, we explain the basic concepts and the architecture of our project and address some key issues related to the project. Finally, we discuss the future directions of our project.

Autonomous navigation is one of primitive functionalities which service robots should have; nevertheless, the navigation problem for a service robot still has many difficulties because the real environment where service robots should works is so complex and dynamical. This paper proposes a framework of intelligent navigation for service robots based on a semantic map of the smart environment. In the smart environment, the robot can receive his position information from location sensors by sensor networks, and it can eliminate the accumulated localization errors. So the robot can accomplish confidential localization in the dynamic and complex environment. With the topological information in the semantic map, path planning problem can be simple even in the wide and complex spaces, and topological path from semantic map can be divided into several sub goals. Robust navigation can be accomplished by moving towards these sub goals with reactive navigation algorithm which has robust characteristics in the dynamic environments. Our approaches have ascertained the good performance on the localization and the navigation, and the feasibility of these methods could be confirmed with the result of experiments in the real environment.

This paper describes a new object recognition system for service robots in a smart environment full of RFID tags and connected by wireless networks. Object recognition is one of the basic functionalities a service robot should perform and many researchers have attempted to make the service robot recognize objects through vision processing in natural environments. However there is no conventional vision system that can recognize target objects robustly in a real-world workspace. In our smart environment, flexibility and robustness in object recognition are provided with the help of RFID tags and communication networks. RFID tags provide the presence and identity information of the object and the robot can recognize the object through vision processing based on a detected RFID code and downloaded visual descriptor information. For feature descriptors, we adopted MPEG-7 visual descriptors due to its concise and unambiguous description of the complex media contents. This paper focuses on developing our object recognition engine based on visual descriptors information in the smart environment. To this end, we propose a new object recognition architecture for service robots, and present a implemented matching algorithm on the basis of MPEG-7 visual descriptors such as color and texture. Experimental results show that the proposed system works well with good performance in terms of recognition rate. This object recognition system will be incorporated into our service robot platform and a pose estimation module is to be included in the next version of our object recognition system.