Unstructured Indoor Environments Research Articles

A novel visual tracking scheme for unstructured indoor environments

In the ever-expanding sphere of assistive robotics, the pressing need for advanced methods capable of accurately tracking individuals within unstructured indoor settings has been magnified. This research endeavours to devise a realtime visual tracking mechanism that encapsulates high performance attributes while maintaining minimal computational requirements. Inspired by the neural processes of the human brain’s visual information handling, our innovative algorithm employs a pattern image, serving as an ephemeral memory, which facilitates the identification of motion within images. This tracking paradigm was subjected to rigorous testing on a Nao humanoid robot, demonstrating noteworthy outcomes in controlled laboratory conditions. The algorithm exhibited a remarkably low false detection rate, less than 4%, and target losses were recorded in merely 12% of instances, thus attesting to its successful operation. Moreover, the algorithm’s capacity to accurately estimate the direct distance to the target further substantiated its high efficacy. These compelling findings serve as a substantial contribution to assistive robotics. The proficient visual tracking methodology proposed herein holds the potential to markedly amplify the competencies of robots operating in dynamic, unstructured indoor settings, and set the foundation for a higher degree of complex interactive tasks.

PILS 시뮬레이터를 이용한 GPS 및 통신불가 미지 실내 환경에서 자율비행 탐색임무를 위한 통합성능 분석

This study presents a PILS simulator design and its performance analysis results of autonomous exploration algorithms under infrastructure outage environments such as battle field. For this, an unstructured indoor environment is considered where drone autonomously explores and returns to the starting point without GPS and other communication linkage. At the end of the mission, the map and the location of the target are transmitted to its operator. Component algorithms necessary for this include path planning, SLAM, image processing with target recognition and its realtime integration for drone’s flight control. The simulator studied in this paper primarily constitutes the flight control system of drone that participates “Autonomous Flight Competition in Battlefield” supported by the Agency for Defense Development(ADD).

Recognition and location of typical automotive parts based on the RGB-D camera

Aiming at the problem that the accuracy of multi-part automatic assembly line sorting is not high, a set of machine vision-based recognition and positioning system is designed with KINECT as the RGB-D camera. The internal and external parameters of the RGB-D camera were calibrated using MATLAB; taking the automobile tire as the target part, because it is better for the system and more accurate, the feature invariant feature transformation (SIFT) algorithm is used to extract and match the feature points of the target part. The depth-based image obtains the spatial position parameters of the target part, thereby calculating the three-dimensional coordinates of the target part, and realizes the recognition and positioning functions of the system. The experimental results show that the visual positioning effectiveness is 96% in the unstructured indoor environment, and the system has good robustness and real-time performance.

Multilevel Humanlike Motion Planning for Mobile Robots in Complex Indoor Environments

In this paper, a multilevel humanlike motion planning approach is proposed for indoor mobile robots. Compared with existing approaches, the novelty of this paper is twofold: 1) the proposed path planning framework is multilevel and humanlike to ensure both foreseeability and flexibility , wherein functions of human brain, eyes, and legs are corresponding to global path planning, sensor-level path planning, and action-level path planning, respectively, and 2) along the planned path, a new velocity-adjustable trajectory planning algorithm is put forward which is provably complete and time optimal considering multiple constraints from both the robot and the environment. Experimental results show that the proposed approach has a better performance in terms of efficiency, smoothness, foreseeability, and flexibility, and autonomous navigation is realized in large-scale, dynamic, partially unknown, and unstructured indoor environments. Note to Practitioners —This paper was motivated by the motion planning problem for mobile robots in complex indoor environments. Existing approaches generally implement a two-level planning scheme, namely, global path planning and local trajectory planning. This paper proposes a novel multilevel humanlike motion planning approach, which is able to guarantee efficiency, smoothness, foreseeability, and flexibility in the presence of kinematic and environmental constraints. Experimental results show that autonomous navigation with the proposed approach is realized in large-scale, dynamic, partially unknown, and unstructured indoor environments, but human motion prediction has not yet been incorporated. In the future research, we will concentrate on human–robot interaction and environment-adaptive planning.

Efficient Office Assistant Robot System: Autonomous Navigation and Controlling Based on ROS

This paper proposes a movable robot that can be used to deliver documents or small parcels (>2.0kg) between employees of an unstructured indoor office environment. The study reviews literature on robot navigation, apply promising techniques and test their accuracy in an unstructured indoor laboratory environment. There are many technologies available in robot navigation. We have selected free and open source software and inexpensive sensors and materials to build a low cost but accurate robot. Our system comprised of three units: first one that process sensor data, second one which is the navigating robot having sensors and actuators, and third one controlling part of office assistant robot system. We have used Robotic Operating System (ROS) indigo as the ground operating system in the background of Ubuntu Operating System. The experiment was conducted to examine the capabilities of the setup and studied its behavior, mapping and localization specifically according to the user commands. An unstructured indoor environment was selected with five users located on the sides. We considered both dark (55 lux – 60 lux) and bright (15 lux – 20 lux) illumination conditions in the room. Then, we allowed the robot to navigate autonomously according to the user commands between five users and monitored its accuracy by looking at the completion of the route to a target user. Results show that the robot has about 98.4% of accuracy of achieving the goal location on average in both lighting conditions and 98% and 98.8% of accuracies for bright and dark conditions respectively. Finally, this office assistant robot system could be practically used in an office environment in both bright and dark lighting conditions.

A Fully-Autonomous Aerial Robot for Search and Rescue Applications in Indoor Environments using Learning-Based Techniques

Search and Rescue (SAR) missions represent an important challenge in the robotics research field as they usually involve exceedingly variable-nature scenarios which require a high-level of autonomy and versatile decision-making capabilities. This challenge becomes even more relevant in the case of aerial robotic platforms owing to their limited payload and computational capabilities. In this paper, we present a fully-autonomous aerial robotic solution, for executing complex SAR missions in unstructured indoor environments. The proposed system is based on the combination of a complete hardware configuration and a flexible system architecture which allows the execution of high-level missions in a fully unsupervised manner (i.e. without human intervention). In order to obtain flexible and versatile behaviors from the proposed aerial robot, several learning-based capabilities have been integrated for target recognition and interaction. The target recognition capability includes a supervised learning classifier based on a computationally-efficient Convolutional Neural Network (CNN) model trained for target/background classification, while the capability to interact with the target for rescue operations introduces a novel Image-Based Visual Servoing (IBVS) algorithm which integrates a recent deep reinforcement learning method named Deep Deterministic Policy Gradients (DDPG). In order to train the aerial robot for performing IBVS tasks, a reinforcement learning framework has been developed, which integrates a deep reinforcement learning agent (e.g. DDPG) with a Gazebo-based simulator for aerial robotics. The proposed system has been validated in a wide range of simulation flights, using Gazebo and PX4 Software-In-The-Loop, and real flights in cluttered indoor environments, demonstrating the versatility of the proposed system in complex SAR missions.

A New Regional Localization Method for Indoor Sound Source Based on Convolutional Neural Networks

At present, the sound source localization methods based on microphone arrays can be roughly classified into three categories: the controllable beamforming technology based on a maximum output power, the high-resolution spectrogram estimation technique, and the sound source localization technique based on time difference of sound. However, an existing localization technology in unstructured indoor environment lacks of localization accuracy and adaptability. In some practical situations, the location of sound source is limited to predefined areas. In this paper, we propose a research method of source region location system based on convolutional neural networks (CNNs). Based on the characteristics of weighted values of CNN, we realize the regional of indoor single sound sources transforming the sound source signals into grammar diagrams and then inputting them into the CNN. The whole process is based on the characteristics of weighted values of CNN. Finally, this paper completes the training and testing for CNN by using the Tensorflow framework. Simulation experiments on the test sets show the effectiveness of the proposed method.

Functional Redesign of Mantis 2.0, a Hybrid Leg-Wheel Robot for Surveillance and Inspection

The paper discusses the redesign of the second version of the Mantis hybrid leg-wheel mobile robot, conceived for surveillance and inspection tasks in unstructured indoor and outdoor environments. This small-scale ground mobile robot is characterized by a main body equipped with two front actuated wheels, a passive rear axle and two rotating legs. Motion on flat and even ground is purely wheeled in order to obtain high speed, high energetic efficiency and stable camera vision; only in case of obstacles or ground irregularities the front legs realize a mixed leg-wheel locomotion to increase the robot climbing ability; in particular, the outer profile of the legs, inspired by the praying mantis, is specially designed to climb square steps. The multibody simulations and the experimental tests on the first prototype have shown the effectiveness of the mixed leg-wheel locomotion not only for step climbing, but also on uneven and yielding terrains. Nevertheless, extensive experimental tests have shown that the front wheels may slip in the last phase of step climbing in case of contact with some materials. In order to overcome this problem, the leg design has been modified with the introduction of auxiliary passive wheels, which reduce friction between legs and step upper surface; these wheels are connected to the legs by one-way bearings, in order to rotate only when they are pulled by the front wheels, and remaining locked when they have to push forward the robot. The influence of the auxiliary wheels on the front wheels slippage is investigated by means of theoretical analysis and multibody simulations.

3-D Depth Reconstruction from a Single Still Image

We consider the task of 3-d depth estimation from a single still image. We take a supervised learning approach to this problem, in which we begin by collecting a training set of monocular images (of unstructured indoor and outdoor environments which include forests, sidewalks, trees, buildings, etc.) and their corresponding ground-truth depthmaps. Then, we apply supervised learning to predict the value of the depthmap as a function of the image. Depth estimation is a challenging problem, since local features alone are insufficient to estimate depth at a point, and one needs to consider the global context of the image. Our model uses a hierarchical, multiscale Markov Random Field (MRF) that incorporates multiscale local- and global-image features, and models the depths and the relation between depths at different points in the image. We show that, even on unstructured scenes, our algorithm is frequently able to recover fairly accurate depthmaps. We further propose a model that incorporates both monocular cues and stereo (triangulation) cues, to obtain significantly more accurate depth estimates than is possible using either monocular or stereo cues alone.

Grid-Based Visual SLAM in Complex Environments

This paper presents a novel approach to the real-time SLAM problem that works in unstructured indoor environment with a single forward viewing camera. Most existing visual SLAM extract features from the environment, associate them in different images and produce a feature map as a result. However, we estimate the distances between the robot and the obstacles by applying a visual sonar ranging technique to the image and then associate this range data through the Iterative Closest Point (ICP) algorithm and finally produce a grid map. Moreover, we construct a pseudo-dense scan (PDS) which is essentially a temporal accumulation of data, emulating a dense omni-directional sensing of the visual sonar readings based on odometry readings in order to overcome the sparseness of the visual sonar and then associate this scan with the previous one. Moreover, we further correct the slight trajectory error incurred in the PDS construction step to obtain a much more refined map using Sequential Quadratic Programming (SQP) which is a well-known optimization scheme. Experimental results show that our method can obtain an accurate grid map using a single camera alone without the need for more expensive.

An optimized modular neural network controller based on environment classification and selective sensor usage for mobile robot reactive navigation

A new approach to the design of a neural network (NN) based navigator is proposed in which the mobile robot travels to a pre-defined goal position safely and efficiently without any prior map of the environment. This navigator can be optimized for any user-defined objective function through the use of an evolutionary algorithm. The motivation of this research is to develop an efficient methodology for general goal-directed navigation in generic indoor environments as opposed to learning specialized primitive behaviors in a limited environment. To this end, a modular NN has been employed to achieve the necessary generalization capability across a variety of indoor environments. Herein, each NN module takes charge of navigating in a specialized local environment, which is the result of decomposing the whole path into a sequence of local paths through clustering of all the possible environments. We verify the efficacy of the proposed algorithm over a variety of both simulated and real unstructured indoor environments using our autonomous mobile robot platform.

Dense 3D Map Construction for Indoor Search and Rescue

AbstractThe main contribution of this paper is a new simultaneous localization and mapping (SLAM) algorithm for building dense three‐dimensional maps using information acquired from a range imager and a conventional camera, for robotic search and rescue in unstructured indoor environments. A key challenge in this scenario is that the robot moves in 6D and no odometry information is available. An extended information filter (EIF) is used to estimate the state vector containing the sequence of camera poses and some selected 3D point features in the environment. Data association is performed using a combination of scale invariant feature transformation (SIFT) feature detection and matching, random sampling consensus (RANSAC), and least square 3D point sets fitting. Experimental results are provided to demonstrate the effectiveness of the techniques developed. © 2007 Wiley Periodicals, Inc.

A Modified Particle Filter for Simultaneous Localization and Mapping

The implementation of a particle filter (PF) for vision-based bearing-only simultaneous localization and mapping (SLAM) of a mobile robot in an unstructured indoor environment is presented in this paper. Variations, using techniques from the genetic algorithm (GA), to standard PF procedures are proposed to alleviate the sample impoverishment problem. A monochrome CCD camera mounted on the robot is used as the measuring device and a measure on the image quality is incorporated into data association and PF update. Since the bearing-only measurement does not contain range information, we add a pseudo range to the measurement during landmark initialization as a hypothesised pair and the non-promising landmark is removed by a map management strategy. Simulation and experimental results from an implementation using real-life data acquired from a Pioneer robot are included to demonstrate the effectiveness of our approach.

Indoor robot navigation using log-polar local maps

In this paper we consider the problem of building a new class of metric maps representing the surroundings of a mobile robot moving in an unstructured indoor environment. Maps are conceived with a metric based on a log-polar space representation: this retina-like representation allows a better definition of objects near the robot, giving less importance to far ones. Information provided at each step by ultrasonic sensors has an uncertainty that is conceptualised as a fuzzy measure and is combined with previous data using Smets transferable believe model. The map building algorithm is integrated in a sensor-based robot navigation system able to recognise some characteristics of the environment. A pattern-matching algorithm, based on Mellin transform, takes advantage on the particular retina-like representation.

Integrating active perception with an autonomous robot architecture

Today’s robotics applications require complex, real-time, high-bandwidth sensor systems. Although many such systems have been developed, integrating them into an autonomous agent architecture remains an area of active research. We have integrated an active stereo vision system with an autonomous agent architecture using a system of perceptual memory. Perceptual memory is an important class of memory because it is designed for the ‘behavior-based’ portion of the agent’s architecture, and not the deliberative portion. This memory maintains current and recent task-dependent perceptual information, as well as expectations about the agent’s immediate environment. Our system of perceptual memory is composed of visual primitives from our stereo system, called proximity spaces. Each proximity space represents a virtual fovea or locus of the agent’s attention. As an application, we present a robot that uses our system of perceptual memory and proximity spaces to ‘attend to’ multiple humans in a complex and unstructured indoor environment.

Cooperative material handling by human and robotic agents: Module development and system synthesis

In this paper we present a collaborative effort to design and implement a cooperative material handling system by a small team of human and robotic agents in an unstructured indoor environment. Our approach makes fundamental use of the human agent's expertise for aspects of task planning, task monitoring and error recovery. Our system is neither fully autonomous nor fully teleoperated. It is designed to make effective use of the human's abilities within the present state of the art of autonomous systems. Our robotic agents refer to systems which are each equipped with at least one sensing modality and which possess some capability for self-orientation and/or mobility. Our robotic agents are not required to be homogeneous with respect to either capabilities or function. Our research stresses both paradigms and testbed experimentation. Theory issues include the requisite coordination principles and techniques which are fundamental to a cooperative multi-agent system's basic functioning. We have constructed an experimental distributed multi-agent architecture testbed facility. The required modular components of this testbed are currently operational and have been tested individually. Our current research focuses on the agent's integration in a scenario for cooperative material handling.