Detection of cracks on bridge decks is a vital task for maintaining the structural health and reliability concrete bridges. Robotic imaging can be used to obtain surface image sets automated on-site analysis. We present novel crack detection algorithm, STRUM (spatially tuned robust multifeature) classifier, demonstrate results real data using state-of-the-art robotic scanning system. By machine learning classification, we eliminate need manually tuning threshold parameters. The algorithm...

For solving the singularity problem arising in control of manipulators, an efficient way is to maximize its manipulability. However, it challenging optimize manipulability effectively because a nonconvex function joint angles robotic arm. In addition, involvement inversion operation expression makes even hard for timely optimization due intensively computational burden matrix inversion. this paper, we make progress on real-time by establishing dynamic neural network recurrent calculation...

Reinforcement learning combined with neural networks has recently led to a wide range of successes in policies different domains. For robot manipulation, reinforcement algorithms bring the hope for machines have human-like abilities by directly dexterous manipulation from raw pixels. In this review paper, we address current status used field. We also cover essential theoretical background and main issues algorithms, which are limiting their applications solving practical problems robotics....

One of the important tasks for bridge maintenance is deck crack inspection. Traditionally, a human inspector detects cracks using his/her eyes and marks location manually. However, accuracy inspection result low due to subjective nature judgement. We propose system that uses camera-equipped mobile robot collect images on deck. In this method, Laplacian Gaussian (LoG) algorithm used detect global map obtained through camera calibration localization. To ensure collects all deck, path planning...

Scalar field mapping has many applications including environmental monitoring, search and rescue, etc. In such applications, there is a need to achieve certain level of confidence regarding the estimates scalar field. this paper, cooperative active sensing framework developed enable using multiple mobile sensor nodes. The controller designed via real-time feedback performance steer sensors new locations in order improve quality. During movement sensors, measurements from each node its...

Multirobot collaboration has great potentials in tasks, such as reconnaissance and surveillance. In this paper, we propose a multirobot system that integrates reinforcement learning flocking control to allow robots learn collaboratively avoid predator/enemy. Our can conduct concurrent distributed fashion well generate efficient combination of high-level behaviors (discrete states actions) low-level (continuous for cooperation. addition, the enables networks how predators while maintaining...

Wild-land fire fighting is a hazardous job. A key task for firefighters to observe the "fire front" chart progress of and areas that will likely spread next. Lack information front causes many accidents. Using unmanned aerial vehicles (UAVs) cover wildfire promising because it can replace humans in tracking significantly reduce operation costs. In this paper, we propose distributed control framework designed team UAVs closely monitor open space, precisely track its development. The UAV team,...

In this paper, autonomous mobile sensor networks are deployed to measure a scalar field and build its map. We develop novel method for multiple nodes map using noisy measurements. Our consists of two parts. First, we distributed fusion algorithm by integrating different consensus filters achieve cooperative sensing among nodes. This has phases. the first phase, weighted average filter is developed, which allows each node find an estimate value at time step. second used allow confidence The...

The condition of bridges is critical for the safety traveling public. Bridges deteriorate with time as a result material aging, excessive loading, environmental effects, and inadequate maintenance. current practice nondestructive evaluation (NDE) bridge decks cannot meet increasing demands highly efficient, cost-effective, safety-guaranteed inspection evaluation. In this paper, mechatronic systems design an autonomous robotic system efficient deck presented. An holonomic mobile robot used...

Abstract The threat to safety of aging bridges has been recognized as a critical concern the general public due poor condition many in United States. Currently, bridge inspection is conducted manually, and it not efficient identify deterioration order facilitate implementation appropriate maintenance or rehabilitation procedures. In this paper, we report new development autonomous mobile robotic system for deck evaluation. robot integrated with several nondestructive evaluation (NDE) sensors...

Wild-land fire fighting is a hazardous job. A key task for firefighters to observe the "fire front" chart progress of and areas it will likely spread next. Lack information front causes many accidents. Using Unmanned Aerial Vehicles (UAV) cover wildfire promising because can replace humans tracking, reducing hazards saving operation costs. In this paper we propose distributed control framework designed team UAVs that closely monitor in open space, precisely track its development. The UAV...

Summary This paper presents visual and 3D structure inspection for steel structures bridges using a developed climbing robot. The robot can move freely on surface, carry sensors, collect data then send to the ground station in real-time monitoring as well further processing. Steel surface image stitching map building are conducted provide current condition of structure. Also, computer vision-based method is implemented detect defects stitched images. effectiveness robot's tested multiple...

Reinforcement learning (RL) enables agents to take decision based on a reward function. However, in the process of learning, choice values for algorithm parameters can significantly impact overall process. In this paper, we use genetic (GA) find used Deep Deterministic Policy Gradient (DDPG) combined with Hindsight Experience Replay (HER), help speed up agent. We method fetch-reach, slide, push, pick and place, door opening robotic manipulation tasks. Our experimental evaluation shows that...

One of the important tasks for bridge maintenance is deck crack inspection. Traditionally, a human inspector detects cracks using his/her eyes and finds location manually. Thus accuracy inspection result low due to subjective nature judgement. We propose system that uses mobile robot conduct inspection, where collects images with high resolution camera. In this method, Laplacian Gaussian algorithm used detect global map obtained through camera calibration localization. To ensure all on deck,...

Formation control of multiple agents has attracted many robotic and researchers recently because its potential applications in various fields. This paper presents a novel approach to the formation obstacle avoidance rectangular with limited communication ranges. The distributed algorithm is designed by utilizing an artificial function. convergence stability analysis proposed given. can guarantee fast performance no collision among agents. Also, proposing repulsive function it as function,...

Ahstract- Unmanned aerial vehicles (UAV) are commonly used for search and rescue missions in unknown environments, where an exact mathematical model of the environment may not be available. This paper proposes a framework UAV to locate missing human after natural disaster such environment, using reinforcement learning (RL) algorithm. A function approximation based RL algorithm is proposed deal with large number states representation obtain faster convergence time. We conducted both simulated...

This paper presents a computer vision-based method to automatically detect concrete cracks. We focus on images containing the concrete: background and crack, where is major mode of gray-scale histogram. Therefore, we address detection problem potential cracks by dealing with histogram thresholding extract regions interests from background. first employ line emphasis moving average filters remove noise surface obtained an inspection robot. The developed algorithm then applied for automatic...

It is necessary for a mobile robot to be able efficiently plan path from its starting, or current location desired goal location. This trivial task when the environment static. However, operational of rarely static, and it often has many moving obstacles. The may encounter one, these unknown unpredictable will need decide how proceed one obstacles obstructing it's path. A method dynamic replanning using RRT* presented. modify an random obstacle obstructs Various experimental results show...

This paper proposes a distributed Multi-Agent Reinforcement Learning (MARL) algorithm for team of Unmanned Aerial Vehicles (UAVs). The proposed MARL allows UAVs to learn cooperatively provide full coverage an unknown field interest while minimizing the overlapping sections among their views. Two challenges in such system are discussed paper: firstly, complex dynamic joint-actions UAV team, that will be solved using game-theoretic correlated equilibrium, and secondly, challenge huge...