Igi Ardiyanto

Real-time navigation using randomized kinodynamic planning with arrival time field

In this paper, we propose a novel path planning algorithm for a mobile robot in dynamic and cluttered environments with kinodynamic constraints. We compute the arrival time field as a bias which gives larger weights for shorter and safer paths toward a goal. We then implement a randomized path search guided by the arrival time field for building the path considering kinematic and dynamic (kinodynamic) constraints of an actual robot. We also consider path quality by adding heuristic constraints on the randomized path search, such as reducing unstable movements of the robot by using a heading criterion. The path will be extracted by backtracking the nodes which reach the goal area to the root of the tree generated by the randomized search, and the motion from the very first node will be sent to the robot controller. We provide a brief comparison between our algorithm and other existing algorithms. Simulation and experimental results prove that our algorithm is fast and reliable to be implemented on the real robot and is able to handle kinodynamic problems effectively.

Partial least squares-based human upper body orientation estimation with combined detection and tracking

This paper deals with the problem of estimating the human upper body orientation. We propose a framework which integrates estimation of the human upper body orientation and the human movements. Our human orientation estimator utilizes a novel approach which hierarchically employs partial least squares-based models of the gradient and texture features, coupled with the random forest classifier. The movement predictions are done by projecting detected persons into 3D coordinates and running an Unscented Kalman Filter-based tracker. The body orientation results are then fused with the movement predictions to build a more robust estimation of the human upper body orientation. We carry out comprehensive experiments and provide comparison results to show the advantages of our system over the other existing methods. We propose a method for estimating the human upper body orientation.Our algorithm uses partial least squares-based gradient and texture feature models.Integration with an UKF-based movement prediction increases the performance.Comparison with the state-of-the-art shows the benefit of our algorithm.Experiment results using image, video, and camera are provided.

Visibility-based viewpoint planning for guard robot using skeletonization and geodesic motion model

This paper describes a viewpoint planning algorithm for a guard robot in an indoor environment. The viewpoint planner is used for the guard robot to watch a certain object such as human continuously. Rather than continuously follows the object, moving the guard robot using the viewpoint planner has many benefits such as reducing the movement and the energy used by the robot. Our viewpoint planner exploits the topology feature of the environment, which is extracted using a skeletonization technique to get a set of viewpoints. We search for escaping gaps from which the target may go out of the robots sight, and make the movement model of the target and the robot to determine the predicted time of the worst case escape of the target. We then plan the action for the robot based on the geodesic model and escaping gaps. Simulation results using 3D simulator are provided to show the effectiveness and feasibility of our algorithm.

Heuristically arrival time field-biased (HeAT) random tree: An online path planning algorithm for mobile robot considering kinodynamic constraints

This paper presents a new path planning algorithm for mobile robot in dynamic environments. We calculate the arrival time field as a bias which gives larger weights for shorter and safer paths towards a goal. We apply a randomized path search guided by the arrival time field for constructing the path considering kinematic and dynamic (kinodynamic) constraints of actual robot. We also consider path quality by adding heuristic constraints such as directing the initial heading of the robot and reducing unstable movements of the robot by using a heading criterion. The path will be extracted by backtracking the nodes which reach the goal area to the root of the tree generated by the randomized search. We provide a brief comparison between our algorithm and other existing algorithms. Simulation and experimental results prove that our algorithm is fast enough to be applied to the real robot and show the effectiveness of the algorithm for handling kinodynamic problems.

Task Oriented Behavior-Based State-Adaptive PID (Proportional Integral Derivative) Control for Low-Cost Mobile Robot

This paper describes how state-adaptive PID (Proportional Integral Derivative) control can be applied to a low-cost mobile robot. Behavior-based state-adaptive control for this mobile robot behaviors was designed using only three infrared sensors, a low-cost 8 bit microcontroller, and an electronic compass, with size of 22cm x 21cm x 16cm. The task oriented behavior-based approach is implemented as two tasks, wall following and goal seeking. Adaptive control used in this robot is PID algorithm using LMS (Least Mean Square) approach. Robot is given a map and run in an artificial corridor representing the map. The results demonstrate that each task works correctly and can run simultaneously. Experimental result shows that robot can run at maximum speed of 100 cm/s without any collision with the corridor. Robot can follow the wall, go to the goal, and avoid obstacles detected by the infrared sensors.

Human motion prediction considering environmental context

This paper describes an approach to predict the human motion. Instead of using a simple motion model as widely used, we take advantages of the environmental context, including the shape and structure, for predicting the human movement. First, we characterize the environment using a graph representation. Subsequently, we acquire the human trajectory tendency on each environment and build a probabilistic sequence model of the human motion. A particle filter-based predictor is then integrated into the system for generating possible future paths of the person. Evaluations on a real campus environment show the advantages of the proposed approach.

3D time-space path planning algorithm in dynamic environment utilizing Arrival Time Field and Heuristically Randomized Tree

This paper deals with a path planning problem in the dynamic and cluttered environments. The presence of moving obstacles and kinodynamic constraints of the robot increases the complexity of path planning problem. We model the environment and motion of dynamic obstacles in 3D time-space. We propose the utilization of the arrival time field for examining the most promising area in those obstacles-occupied 3D time-space for approaching the goal. The arrival time field is used for guiding the expansion of a randomized tree search in a favorable way, considering kinodynamic constraints of the robot. The quality and the optimality of the path are taken into account by performing heuristic methods on the randomized tree. Simulation results are also provided to prove the feasibility, possibility, and effectiveness of our algorithm.

Deep residual coalesced convolutional network for efficient semantic road segmentation

This paper proposes a deep learning-based efficient and compact solution for road scene segmentation problem, named deep residual coalesced convolutional network (RCC-Net). Initially, the RCC-Net performs dimensionality reduction to compress and extract relevant features, from which it is subsequently delivered to the encoder. The encoder adopts the residual network style for efficient model size. In the core of each residual network, three different convolutional layers are simultaneously coalesced for obtaining broader information. The decoder is then altered to upsample the encoder for pixel-wise mapping from the input images to the segmented output. Experimental results reveal the efficacy of the proposed network over the state-of-the-art methods and its capability to be deployed in an average system.

Maximum entropy principle for exudates segmentation in retinal fundus images

This paper addresses a novel segmentation algorithm for detecting one of the diabetic retinopathy pathologies, called “exudates”. Exudates segmentation is ordinarily examined from retinal fundus images by various image processing techniques. Instead of carefully picking up the specific exudates features on the retinal images as has been done by the other works, our scheme is to observe global information of the retinal images. The global information, as well as spatial information, is extracted by maximum entropy-based thresholding. The proposed algorithm determines a reasonable threshold value for separating exudates areas, which are usually sparse and brighter, from the rest of images. This approach also ensures and minimizes the illumination variance effects of different images since it takes into account the global information. In addition to the proposed algorithm, luminance channel of the retinal images is exploited for pre-processing stage. After the optical disc which has similar characteristic to the exudates is separated, the pathological areas are subsequently acquired. Evaluations on the E-OPHTHA-EX retinal fundus images database show the advantages of the proposed approach, with the accuracy 99.4 percent, specificity 99.6 percent, and sensitivity 16.9 percent.

Classification of breast ultrasound images based on texture analysis

Ultrasonography (USG) is a popular imaging modality because of its flexibility, non-invasion, non-ionisation and low cost. A breast ultrasound used to detect and classify abnormalities of the breast mass. However, the diagnosis is very subjective because it depends on the ability of the radiologist. In order to eliminate operator dependency and to improve the diagnostic accuracy, a computerised system is necessary to do the feature extraction and the classification of the breast nodule. This research proposes a classification of breast USG images by using some texture features into two classes. The dataset consists of 57 USG images which grouped into 27 anechoic cases and 30 hypoechoic cases. An initial step of image pre-processing is conducted to enhance the detection capability. Afterwards, followed by some methods of morphological operation, region growing active contour and histogram equalization. The feature extraction method used texture analysis, which is histogram, gray level co-occurrence matrix (GLCM) and fractal Brownian motion (FBM). Finally, Multilayer Perceptron (MLP) classification method is used to classify anechoic nodule from hypoechoic nodule. The result shows that the proposed method achieved the accuracy of 91.23%, sensitivity of 95.83%, specificity of 87.88%, Positive Predictive Value (PPV) of 85.19% and Negative Predictive Value (NPV) of 96.67%. This suggest that the proposed method is excellent in analyzing breast USG images.