Robot Navigation Tasks Research Articles

Path Planning Methods in an Environment with Obstacles (A Review)

Planning the path is the most important task in the mobile robot navigation. This task involves basically three aspects. First, the planned path must run from a given starting point to a given endpoint. Secondly, it should ensure robot’s collision-free movement. Thirdly, among all the possible paths that meet the first two requirements it must be, in a certain sense, optimal.Methods of path planning can be classified according to different characteristics. In the context of using intelligent technologies, they can be divided into traditional methods and heuristic ones. By the nature of the environment, it is possible to divide planning methods into planning methods in a static environment and in a dynamic one (it should be noted, however, that a static environment is rare). Methods can also be divided according to the completeness of information about the environment, namely methods with complete information (in this case the issue is a global path planning) and methods with incomplete information (usually, this refers to the situational awareness in the immediate vicinity of the robot, in this case it is a local path planning). Note that incomplete information about the environment can be a consequence of the changing environment, i.e. in a dynamic environment, there is, usually, a local path planning.Literature offers a great deal of methods for path planning where various heuristic techniques are used, which, as a rule, result from the denotative meaning of the problem being solved. This review discusses the main approaches to the problem solution. Here we can distinguish five classes of basic methods: graph-based methods, methods based on cell decomposition, use of potential fields, optimization methods, фтв methods based on intelligent technologies.Many methods of path planning, as a result, give a chain of reference points (waypoints) connecting the beginning and end of the path. This should be seen as an intermediate result. The problem to route the reference points along the constructed chain arises. It is called the task of smoothing the path, and the review addresses this problem as well.

Fuzzy Logic-Ant Colony Optimization for Explorer-Follower Robot with Global Optimal Path Planning

Path planning is an essential task for the mobile robot navigation. However, such a task is difficult to solve, due to the optimal path needs to be rerouted in real-time when a new obstacle appears. It produces a sub-optimal path and the robot can be trapped in local minima. To overcome the problem the Ant Colony Optimization (ACO) is combined with Fuzzy Logic Approach to make a globally optimal path. The Fuzzy-ACO algorithm is selected because the fuzzy logic has good performance in imprecision and uncertain environment and the ACO produce simple optimization with an ability to find the globally optimal path. Moreover, many optimization algorithms addressed only at the simulation level. In this research, the real experiment is conducted with the low-cost Explorer-Follower robot. The results show that the proposed algorithm, enables them to successfully identify the shortest path without collision and stack in “local minima”.

Laser variational autoencoder for map construction and self-localization

For accurate global self-localization with small memory usage, researches for the compression of the laser-scan data have been actively conducted. Main approaches to the compression are to design feature extractor based on human knowledge regarding the specific environment, e.g., office and hallway. However, in real robot navigation tasks such as a security patrol robot, the robot would be applied to a variety of environments and it is expensive if the users need to tune the design at every environment. To alleviate such problem, we propose to extend the state-of-the-art variational auto-encoder (VAE) by introducing the step-edge detector, which detects non-continuous transition emerged frequently at the laser scan data due to the limitation of distance measurement. With our proposed method, called laserVAE, the feature extractor of the laser scan is automatically tuned given unknown environments. Through experiments with a real self-localization with 2D laser scan, we demonstrate the effectiveness of the proposed method.

A Method of Mobile Robot Position Estimation Correction Using Visual Location of Natural Landmarks

The article presents an approach to using of visual location of an arbitrary number of known natural landmarks for correction of position estimation of a mobile robot. The approach is based on using of the Extended Kalman Filter to perform steps of the position prediction and correction based on the visual location. The visual location is performed using a calibrated camera (or a set of cameras) installed on the mobile robot. The task of a mobile robot navigation in a case of poorly determined conditions is often soled using a set of expensive LIDARs. Other sensors like GNSS and odometer often used in mobile robots are usually not precise enough (for example when maneuvering on intersections). The recent research in the field of computer vision allows creation of much less expensive systems based on image analysis from one or several cameras. When used together with other sensors such system can significantly increase the navigation precision and stability. Known visual navigation approaches like visual SLAM and visual odometer are often used but they are often not precise enough especially when the camera movement is mostly rotating. A set of one or more natural visual landmarks can be located automatically in real time using mobile robot cameras. Existing methods provide partial solutions for the position or direction estimation using 3 or more landmarks. The proposed approach is based on using of the Extended Kalman Filter for efficient fusion of the odometer and GNSS data with the visual location of one or more landmarks. Since all the landmarks are used one at a time independently the number of landmarks, cameras, their types and positions are arbitrary. The method is working with one camera and one landmark as well as with a round view camera system with one or many landmarks. Experiments involving autonomous driving through different intersections shown the feasibility of reaching of the accuracy of horizontal position of 10 cm and rotation of 1° and below. The method has been developed for improving of unmanned vehicles navigation at intersections but can be applied to navigation of different ground, space, marine and underwater mobile robots.

A navigation strategy for an autonomous patrol vehicle based on multi-fusion planning algorithms and multi-paradigm representation schemes

This paper considers the robot navigation task as an algorithmic and representational one in a way that the performance of the navigation task cannot be measured without combining those two elements. As a result of this view, a unique new navigation strategy based on combined multi-fusion planning algorithms and multi-paradigm representation schemes is presented. An overall architecture for a new navigation strategy is also proposed. Discrete and continuous planning algorithms are combined in a hierarchal fashion. GIS models and ontology are also combined to form rich media for representing dynamic data and knowledge. Experimental results with an evaluation of the schemes are presented.

Cyclic error correction based Q-learning for mobile robots navigation

Similar to control systems, reinforcement learning can capture notions of optimal behavior using natural interaction experience. In the context of reinforcement learning, the temporal difference error of the generated experience measures how well the learner responds to the system. Specially sequential difference of accumulated temporal difference error can indicate the learning performance. In this paper, we fully utilize the error correction in closed-loop peculiarity by mapping a representation error to the step-size component. The proposed cyclic step-size could better control how new estimates are iteratively blended together over time, and the new estimates guide the action selection process which in turn influence the value distribution. To guide more promising action decision, an ensemble action selector is proposed which incorporates the idea of ensemble wisdom of the weak. Experimental results conducted under gridworld mobile robot navigation task demonstrate the validity, capacity of fast learning and easy-plugged implementation of the derived algorithm, leading to increasing applicability to real-life problems.

A Fast Ellipse Detector Using Projective Invariant Pruning.

Detecting elliptical objects from an image is a central task in robot navigation and industrial diagnosis, where the detection time is always a critical issue. Existing methods are hardly applicable to these real-time scenarios of limited hardware resource due to the huge number of fragment candidates (edges or arcs) for fitting ellipse equations. In this paper, we present a fast algorithm detecting ellipses with high accuracy. The algorithm leverages a newly developed projective invariant to significantly prune the undesired candidates and to pick out elliptical ones. The invariant is able to reflect the intrinsic geometry of a planar curve, giving the value of -1 on any three collinear points and +1 for any six points on an ellipse. Thus, we apply the pruning and picking by simply comparing these binary values. Moreover, the calculation of the invariant only involves the determinant of a 3×3 matrix. Extensive experiments on three challenging data sets with 648 images demonstrate that our detector runs 20%-50% faster than the state-of-the-art algorithms with the comparable or higher precision.

The Emotional, Cognitive, Physiological, and Performance Effects of Variable Time Delay in Robotic Teleoperation

The effects of intermittent and variable time delay were investigated to understand the cognitive and physical consequences of gaps between an input from an operator and the corresponding feedback response from the system. Time delay has been shown to disrupt task performance in various areas including psychology and telerobotics. Previous research in multiple domains has focused on the performance effects of time delay and overcoming technological limitations that cause time delay. However, robotics researchers have yet to study the effects of variable time delay on specific operator emotions, usability, and physiological activation in teleoperations. This study investigates the influence of variable time delay not only on task performance, but also operator emotions, physiological arousal, cognitive workload, and usability in teleoperation. Time delay was manipulated by introducing lag into the system feedback. Participants were asked to navigate a remote-control robot vehicle through mazes of differing levels of task complexity in a remote location and simultaneously identify targets. Results showed that operator frustration, anger, and workload increased while usability and task performance decreased when intermittent and variable feedback lag was introduced to a robotic navigation task. The effect of the variable time delay was greater than the effect of task complexity. Furthermore, results suggest that the effects are of time delay and task complexity are additive. A better understanding of the emotional experiences of human operators and the corresponding physiological signals is of crucial importance to designing affect-aware robotic systems that have the ability to appropriately mitigate negative operator emotional states.

Improved Object Detection Using a Robotic Sensing Antenna with Vibration Damping Control

Some insects or mammals use antennae or whiskers to detect by the sense of touch obstacles or recognize objects in environments in which other senses like vision cannot work. Artificial flexible antennae can be used in robotics to mimic this sense of touch in these recognition tasks. We have designed and built a two-degree of freedom (2DOF) flexible antenna sensor device to perform robot navigation tasks. This device is composed of a flexible beam, two servomotors that drive the beam and a load cell sensor that detects the contact of the beam with an object. It is found that the efficiency of such a device strongly depends on the speed and accuracy achieved by the antenna positioning system. These issues are severely impaired by the vibrations that appear in the antenna during its movement. However, these antennae are usually moved without taking care of these undesired vibrations. This article proposes a new closed-loop control schema that cancels vibrations and improves the free movements of the antenna. Moreover, algorithms to estimate the 3D beam position and the instant and point of contact with an object are proposed. Experiments are reported that illustrate the efficiency of these proposed algorithms and the improvements achieved in object detection tasks using a control system that cancels beam vibrations.

ROS プラットフォームにおける移動ロボットの自律走行のための 動作獲得学習および走行シミュレーション

This paper describes a system for learning and autonomous navigation of mobile robots on ROS (Robot Operating System) platform. Applying learning of autonomous navigation in the ROS platform will increase availability for autonomous navigation in actual environments, because the ROS platform is often utilized for navigation system in actual environments such as Tsukuba Challenge. Especially, in this study, a learning system based on Deep Q-Networks, that is effective for learning tasks in high-dimensional state spaces, is introduces. At first, simple maze problems as simplified robot navigation environment are solved. Then, the maze problems are applied to simulated robot navigation tasks on ROS platform. Learning results show that the robot obtained a policy to reach any destinations in the maps of the ROS simulator.

Path planning of multi-agent systems in unknown environment with neural kernel smoothing and reinforcement learning

Path planning is a basic task of robot navigation, especially for autonomous robots. It is more complex and difficult for multi-agent systems. The popular reinforcement learning method cannot solve the path planning problem directly in unknown environment.In this paper, the classical multi-agent reinforcement learning algorithm is modified such that it does not need the unvisited state. The neural networks and kernel smoothing techniques are applied to approximate greedy actions by estimating the unknown environment. Experimental and simulation results show that the proposed algorithms can generate paths in unknown environment for multiple agents.

From free energy to expected energy: Improving energy-based value function approximation in reinforcement learning

Free-energy based reinforcement learning (FERL) was proposed for learning in high-dimensional state and action spaces. However, the FERL method does only really work well with binary, or close to binary, state input, where the number of active states is fewer than the number of non-active states. In the FERL method, the value function is approximated by the negative free energy of a restricted Boltzmann machine (RBM). In our earlier study, we demonstrated that the performance and the robustness of the FERL method can be improved by scaling the free energy by a constant that is related to the size of network. In this study, we propose that RBM function approximation can be further improved by approximating the value function by the negative expected energy (EERL), instead of the negative free energy, as well as being able to handle continuous state input. We validate our proposed method by demonstrating that EERL: (1) outperforms FERL, as well as standard neural network and linear function approximation, for three versions of a gridworld task with high-dimensional image state input; (2) achieves new state-of-the-art results in stochastic SZ-Tetris in both model-free and model-based learning settings; and (3) significantly outperforms FERL and standard neural network function approximation for a robot navigation task with raw and noisy RGB images as state input and a large number of actions.

Computationally efficient visual–inertial sensor fusion for Global Positioning System–denied navigation on a small quadrotor

Because of the complementary nature of visual and inertial sensors, the combination of both is able to provide fast and accurate 6 degree-of-freedom state estimation, which is the fundamental requirement for robotic (especially, unmanned aerial vehicle) navigation tasks in Global Positioning System–denied environments. This article presents a computationally efficient visual–inertial fusion algorithm, by separating orientation fusion from the position fusion process. The algorithm is designed to perform 6 degree-of-freedom state estimation, based on a gyroscope, an accelerometer and a monocular visual-based simultaneous localisation and mapping algorithm measurement. It also recovers the visual scale for the monocular visual-based simultaneous localisation and mapping. In particular, the fusion algorithm treats the orientation fusion and position fusion as two separate processes, where the orientation fusion is based on a very efficient gradient descent algorithm, whereas the position fusion is based on a 13-state linear Kalman filter. The elimination of the magnetometer sensor avoids the problem of magnetic distortion, which makes it a power-on-and-go system once the accelerometer is factory calibrated. The resulting algorithm shows a significant computational reduction over the conventional extended Kalman filter, with competitive accuracy. Moreover, the separation between orientation and position fusion processes enables the algorithm to be easily implemented onto two individual hardware elements and thus allows the two fusion processes to be executed concurrently.

Goal-directed affordance prediction at the subtask level

Purpose – This paper aims to present a hybrid control approach that combines learning-based reactive control and affordance-based deliberate control for autonomous mobile robot navigation. Unlike many current navigation approaches which only use learning-based paradigms, the authors focus on how to utilize the machine learning methods for reactive control together with the affordance knowledge that is simultaneously inherent in natural environments to gain advantages from both local and global optimization. Design/methodology/approach – The idea is to decompose the complex and large-scale robot navigation task into multiple sub-tasks and use the hierarchical reinforcement learning (HRL) algorithm, which is well-studied in the learning and control algorithm domains, to decompose the overall task into sub-tasks and learn a grid-topological map of the environment. An affordance-based deliberate controller is used to inspect the affordance knowledge of the obstacles in the environment. The hybrid control architecture is then designed to integrate the learning-based reactive control and affordance-based deliberate control based on the grid-topological and affordance knowledge. Findings – Experiments with computer simulation and an actual humanoid NAO robot have demonstrated the effectiveness of the proposed hybrid approach for mobile robot navigation. Originality/value – The main contributions of this paper are a new robot navigation framework that decomposes a complex navigation task into multiple sub-tasks using the HRL approach, and hybrid control architecture development that integrates learning-based and affordance-based paradigms for autonomous mobile robot navigation.

Accurate 3-D Position and Orientation Method for Indoor Mobile Robot Navigation Based on Photoelectric Scanning

Position and orientation of indoor mobile robots must be obtained real timely during operation in structured industrial environment, so as to ensure the security and efficiency of cargo transportation and assembly precision. But for such a large-scale space, only 2-D coordinates and heading angle of the mobile robot can be measured with a relatively low precision in current major methods. This paper presents a novel method for 3-D position and orientation measurement of indoor mobile robot. In this method, a rotary-laser transmitter is utilized, which is mounted on the indoor mobile robot measuring the scanning angles relative to photoelectric artificial landmarks and obtaining its own 3-D space location information. The landmarks whose coordinates in navigation frame should be precalibrated are distributed at the most appropriate positions of the structured industrial environment. On the basis of that, an algorithm of multiangle intersection was established and in-depth discussed to solve transmitter’s spatial position and orientation. Experimental results show that, in an $8~ {\rm m}\times 6~{\rm m}\times 2.5$ m working volume, transmitter’s position, and orientation measurement accuracy of proposed method were higher than 3.8 mm and 0.104°, respectively. It demonstrates that the proposed method is reliable and flexible for indoor mobile robot navigation tasks and the measurement accuracy can be further improved by increasing layout density of landmarks.

A Cognitive Architecture Based on a Learning Classifier System with Spiking Classifiers

Learning classifier systems (LCS) are population-based reinforcement learners that were originally designed to model various cognitive phenomena. This paper presents an explicitly cognitive LCS by using spiking neural networks as classifiers, providing each classifier with a measure of temporal dynamism. We employ a constructivist model of growth of both neurons and synaptic connections, which permits a genetic algorithm to automatically evolve sufficiently-complex neural structures. The spiking classifiers are coupled with a temporally-sensitive reinforcement learning algorithm, which allows the system to perform temporal state decomposition by appropriately rewarding “macro-actions”, created by chaining together multiple atomic actions. The combination of temporal reinforcement learning and neural information processing is shown to outperform benchmark neural classifier systems, and successfully solve a robotic navigation task.

Signal processing for velocity and range measurement using a micromachined ultrasound transducer

Signal processing techniques are under investigation for determination of range and velocity information from MEMS based ultrasound transducers. The ideal technique will be real-time, result in high resolution and accurate measurements, and operate successfully in noise. Doppler velocity measurements were previously demonstrated using a MEMS cMUT array (Shin et al., ASA Fall Meeting 2011, JASA 2013, Sens. Actuators A 2014). The MEMS array has 168 nickel-on-glass capacitive ultrasound transducers on a 1 cm die, and operates at 180 kHz in air. Post processing of the received ultrasound demonstrated the ability to sense velocity using continuous wave (CW) Doppler at a range of up to 1.5 m. The first attempt at real-time processing using a frequency modulated continuous wave (FM/CW) scheme was noise limited by the analog demodulation circuit. Further noise analysis is ongoing to determine whether this scheme may be viable. Other schemes under consideration include cross correlation chirp and single and multi-frequency burst waveforms. Preliminary results from a single frequency burst showed that cross-correlation-based signal processing may achieve acceptable range. The system is targeted at short range small robot navigation tasks. Determination of surface roughness from scattering of the reflected waves may also be possible.

Mobile Robot Navigation and Obstacle-avoidance using ANFIS in Unknown Environment

Navigation and obstacle avoidance in an unknown environment is proposed in this paper using hybrid neural network with fuzzy logic controller. The overall system is termed as Adaptive Neuro Fuzzy Inference System (ANFIS). ANFIS combines the benefits of fuzzy logic and neural networks for the purpose of achieving robotic navigation task. Simulation results are presented using Khepera Simulator (KiKs) within MATLAB environment. Moreover, experimental results are obtained using Khepera III platform.

Towards A Human Robot Interaction Framework with Marker-less Augmented Reality and Visual SLAM

This paper presents a novel framework for Human Robot Interaction (HRI) using marker-less Augmented Reality (AR). Unlike marker-based AR, marker-less AR does not require the environment to be instrumented with special markers and so it works favorably for unknown/unprepared environments. Current state-of-the-art visual SLAM approaches like PTAMM (Parallel Tracking and Multiple Mapping) achieve this with constrained motion models within local co-ordinate systems. Our framework relaxes motion model constraints enabling a wider range of camera movements to be robustly tracked and extends PTAMM with a series of linear transformations. The linear transformations enable AR markers to be seamlessly placed and tracked within a global co-ordinate system of any size. This allows us to place markers globally and view them from any direction and perspective, even when returning to the markers from a different direction or perspective. We report on the model's performance and show how the model can be applied to help humans interact with robots. In this paper we look at how they can assist robot navigation tasks. 

Comparing Action as Input and Action as Output in a Reinforcement Learning Task

Generalization techniques are useful for enabling an agent to be able to approximate the value of states it has not encountered so far in reinforcement learning. They are also useful as memory use minimization mechanisms in situations where the state space is too large such that is infeasible to represent every state in the state space in the computer memory. Artificial Neural Networks are one generalization technique that is usually employed. Various network structures have been proposed in literature. In this study, two of the structures that have been proposed were implemented in a robot navigation task and their performance compared. The results indicate that having a network structure where there is an output node for each of the possible actions, is superior to the structure in which the selected action is fed as an input to the network and its value output by the single network output node. General Terms