ICARSC 2019: 19TH IEEE INTERNATIONAL CONFERENCE ON AUTONOMOUS ROBOT SYSTEMS AND COMPETITIONS
PROGRAM FOR THURSDAY, APRIL 25TH
Days:
previous day
next day
all days

View: session overviewtalk overview

09:00-10:00 Session 8: Localization
09:00
Parallelization of a Vine Trunk Detection Algorithm For a Real Time Robot Localization System

ABSTRACT. Developing ground robots for crop monitoring and harvesting in steep slope vineyards is a complex challenge due to two main reasons: harsh condition of the terrain and unstable localization accuracy obtained with Global Navigation Satellite System (GNSS). In this context, a reliable localization system requires an accurate detector for high density of natural/artificial features. In previous works, we presented a novel visual detector for Vineyards Trunks and Masts (ViTruDe) with high levels of detection accuracy. However, its implementation on the most common processing units - central processing units (CPU), using a standard programming language (C/C++), is unable to reach the processing efficiency requirements for real time operation. In this work, we explored parallelization capabilities of processing units, such as graphics processing units (GPU), in order to accel- erate the processing time of ViTruDe. This work gives a general perspective on how to parallelize a generic problem in a GPU based solution, while exploring its efficiency when applied to the problem at hands. The ViTruDe detector for GPU was developed considering the constraints of a cost-effective robot to carry-out crop monitoring tasks in steep slope vineyard environments. We compared the proposed ViTruDe implementation on GPU and CPU, and the achieved solution is over eighty times faster than its CPU counterpart. The training and test data are made public for future research work. This approach is a contribution for an accurate and reliable localization system that is GNSS-free.

09:15
New Approach to Supervise Localization Algorithms

ABSTRACT. The localization algorithms have different errors which can impair the robot’s navigation. In this way, we propose an approach that will supervise the localization while the robot navigate. Our approach is based on another work present in the literature, where we detected a problem during its analysis. Therefore, this article will present a new method based on the RLS algorithm, to solve the identified problem. Besides, we propose the supervision of two more localization algorithms, being now four the supervised algorithms, namely: Augmented Monte Carlo Localization, Extended Kalman Filter with Beacons, Perfect Match and Odometry. The results show that the robustness and reliability of the system were increased.

09:30
Robot localization Through Optical Character Recognition of Signs

ABSTRACT. Optical character recognition is the process by which the textual content of an image is converted into strings. Localization is the problem of figuring out where one is in a given environment. In this work we approach the application of optical character recognition (OCR) in robot localization. We propose, develop and test a vision based localization system that is capable of detecting room identification signs present in the environment, recognizing their textual contents and apply them to determine its position referent to a topological map of the environment. A method to detect the signs based on image segmentation by color and corners detection by the analysis of its contour is developed. The recognition of characters is performed with the application of an open-source OCR engine. Localization is realized through the comparison of sign readings with the textual information embedded in the topological representation of the environment. The algorithm was tested in a dataset of images acquired in a corridor. The experimental results show that the developed system is capable of successfully determining its localization in 54,54% of tested cases.

09:45
A performance comparison of bio-inspired behaviours for odour source localisation

ABSTRACT. Detecting and locating odour sources with mobile robots is hard, yet an interesting problem for many real world applications. In nature, this problem is daily and successfully addressed as a way to survive. This success motivated researchers to adapt the observed biological behaviours into robotic search strategies. Each of these strategies are meant to operate under specific environmental conditions and, despite being inherently different, they can be decomposed into a small set of behaviours. The present paper compares the performance of those behaviours in finding and tracking odour plumes to their sources, under diverse environmental conditions. The experimental results show that the performance of each behaviour is highly dependent on the environmental conditions and thus the behaviour employed must be carefully selected. The resulting knowledge may be used by the community for producing better performing strategies, either through hand-design or learning methods.

10:10-11:00 Session 9: keynote 2

Robot Competitions Driving Robotics Research

Pedro U. Lima

11:20-12:05 Session 10: Learning and Detection
11:20
Comparing Spatial and Mobile Augmented Reality for Guiding Assembling Procedures with Task Validation

ABSTRACT. Assembly tasks are a common situation in many industrial applications. These tasks are often presented on paper or digital manuals containing instructions, photos or diagrams to guide an assembly sequence. While some Augmented Reality (AR) systems have also been proposed to support these processes, only a few track the state of the assembling procedure, validating the process in real-time. In this work, we propose two different AR-based (mobile and spatial AR) methods with real-time validation to provide assistance to users during the execution of an assembly process. The validation process uses computer vision techniques to keep track of the state of the assembly sequence, verifying the completion of each stage and providing information at the end of the assembly. A controlled experiment was used to compare the performance, ease of use, and acceptance of the two AR-based methods proposed. Participants were significantly faster and made fewer errors using the Spatial AR condition. Besides, participants also preferred this condition. In addition, Nasa TLX rating showed that the Spatial AR condition proven to have a slightly lower cognitive load on the participants.

11:35
Detecting Robotic Anomalies using RobotChain

ABSTRACT. Robotic events can provide notable amounts of information regarding a robot's status, which can be extrapolated to detect productivity, anomalies, malfunctions and used for monitorization. However, when problems occur in sensitive environments like a factory, the logs of a machine may be discarded because they are susceptible to chances and malicious intents. In this paper we propose to use RobotChain for anomaly detection. RobotChain is a method to securely register robotic events, using a blockchain, which ensures that once an event gets registered on it, it's secured and cannot be tampered with. We show how this system can be leveraged with the module for anomaly detection, that uses the information contained on the blockchain to detect anomalies on a UR3 robot.

11:50
Feature extraction using Hidden Markov Models for 3D-visual recognition

ABSTRACT. In this work, we present a novel implementation for visual recognition using probabilistic models. Given a scene view, we first propose a 3D feature extraction from a point cloud of map features as a series of observations in a Hidden Markov Model; then, we evaluate the Profile HMM in the place recognition task using a publicly available dataset. Furthermore, we evaluated a classical HMM in the object recognition task in the context of anthropomorphic service robots. Results show that our approach performs well in the aforementioned tasks with high recognition rates.

12:15-13:00 Session 11: Autonomous Vehicles
12:15
Real-Time LiDAR-based Power Lines Detection for Unmanned Aerial Vehicles

ABSTRACT. The growing dependence of modern-day societies on electricity increases the importance of effective monitoring and maintenance of power lines. Endowing UAVs with the appropriate sensors for inspecting power lines, the costs and risks associated with the traditional foot patrol and helicopter-based inspections can be reduced. However, this implies the development of algorithms to make the inspection process reliable and autonomous. Visual methods are usually applied to locate the power lines and their components, but poor light conditions or a background rich in edges may compromise their results. To overcome those limitations, we propose to address the problem of power line detection and modeling based on LiDAR. A novel approach to the power line detection was developed, PL2DM - Power Line LiDAR-based Detection and Modeling. It is based in a scan-by-scan adaptive neighbor minimalist comparison for all the points in a point cloud. The power line final model is obtained by matching and grouping several line segments, using their collinearity properties. Horizontally, the power lines are modeled as a straight line, and vertically as a catenary curve. The algorithm was validated with a real dataset, showing promising results both in terms of outputs and processing time, adding real-time object-based perception capabilities for other layers of processing.

12:30
Hybrid Approach to Estimate a Collision-Free Velocity for Autonomous Surface Vehicles
SPEAKER: Renato Silva

ABSTRACT. Shipping transportation mode needs to be even more efficient, profitable and secure as more than 80% of the world’s trade is done by sea. Autonomous ships will provide the possibility to eliminate the likelihood of human error, reduce unnecessary crew costs and increase the efficiency of the cargo spaces. Although a significant work is being made, and new algorithms are arising, they are still a mirage and still have some problems regarding safety, autonomy and reliability. This paper proposes an online obstacle avoidance algorithm for Autonomous Surfaces Vehicles (ASVs) introducing the reachability with the protective zone concepts. This method estimates a collision-free velocity based on inner and outer constraints such as, current velocity, direction, maximum speed and turning radius of the vehicle, position and dimensions of the surround obstacles as well as a movement prediction in a close future. A non-restrictive estimative for the speed and direction of the ASV is calculated by mapping a conflict zone, determined by the course of the vehicle and the distance to obstacles that is used to avoid imminent dangerous situations. A set of experiments demonstrates the ability of this method to safely circumvent obstacles in several scenarios with different weather conditions.

12:45
An Hierarchical Architecture for Docking Autonomous Surface Vehicles
SPEAKER: Pedro Leite

ABSTRACT. Autonomous Surface Vehicles (ASVs) provide the ideal platform to further explore the many opportunities in the cargo shipping industry, by making it more profitable and safer. This paper presents an architecture for the autonomous docking operation, formed by two stages: a maneuver module and, a situational awareness system to detect a mooring facility where an ASV can safely dock. Information retrieved from a 3D LIDAR, IMU and GPS are combined to extract the geometric features of the floating platform and to estimate the relative positioning and orientation of the moor to the ASV. Then, the maneuver module plans a trajectory to a specific position and guarantees that the ASV will not collide with the mooring facility. The approach presented in this paper was validated in distinct environmental and weather conditions such as tidal waves and wind. The results demonstrate the ability of the proposed architecture for detecting the docking platform and safely conduct the navigation towards it, achieving errors up to 0.107 m in position and 0.115 rads in orientation.

14:00-15:00 Session 12: Modeling and Programming
14:00
Converting Robot Offline Programs to Native Code Using the AdaptPack Studio Translators

ABSTRACT. The increase in productivity is a demand for modern industries that need to be competitive in the actual business scenario. To face these challenges, companies are increasingly using robotic systems for end-of-line production tasks, such as wrapping and palletizing, as a mean to enhance the production line efficiency and products traceability, allowing human operators to be moved to more added value operations. Despite this increasing use of robotic systems, these equipments still present some inconveniences regarding the programming procedure, as the time required for its execution does not meet the current industrial needs. To face this drawback, offline robot programming methods are gaining great visibility, as their flexibility and programming speed allows companies to face the need of successive changes in the production line set-up. However, even with a great number of robots and simulators, the efforts to merge them did not reach the needs of engineers. Therefore, this paper proposes a translation library named AdaptPack Studio Translator, which is capable to export proprietary codes for the ABB, Fanuc, Kuka, and Yaskawa robot brands, after their offline programming has been performed in the Visual Components software. The results presented in this paper are evaluated in simulated and real scenarios.

14:15
A GSPN Software Framework to \\ Model and Analyze Robot Tasks

ABSTRACT. In this paper we introduce a software framework to represent robot task plans based on generalized stochastic Petri nets. Our framework allows modeling and analysis of a robot task, providing structural and performance metrics of the designed Petri net, making it a systematic design-analysis-design tool, that leads to improved task plans before execution in real robots. Results of a case study with multiple robots in a virtual scenario show the ability of the framework to provide metrics and relevant properties of the designed task, and how to quickly optimize it.

14:30
AdaptPack Studio: Automatic Offline Robot Programming Framework for Factory Environments

ABSTRACT. The brisk and dynamic environment that factories are facing, both as an internal and an external level, requires a collection of dexterous tools to solve emerging issues in the industry 4.0 context. Part of the common challenges that appear are related to the increasing demand for high adaptability in the organizations’ production lines. Mechanical processes are becoming faster and more adjustable to the production diversity in the Fast Moving Consumer Goods (FMCG). Concerning the previous characteristics, future factories can only remain competitive and profitable if they have the ability to quickly adapt all their production resources in response to inconstant market demands. Having previous concerns in focus, this paper presents a fast and adaptative framework for automated cells modeling, simulation and offline robot programming, focused on palletization operations. Established as an add-on for the Visual Components 3D manufacturing simulation software, the proposed application allows performing fast layout modeling and selection of pick and place points. Furthermore, A* based algorithms are used for generating collision-free trajectories, discretized both in the robot joint space and in the Cartesian space. The software evaluation was tested inside the Visual Components simulation world, which models real factory environments. Results have shown to be concise and accurate, with minor displacement inaccuracies due to differences between the virtual model and the real world.

14:45
Shop Floor Virtualization and Industry 4.0

ABSTRACT. This paper addresses one of the key components in today's industrialization approach: virtualization. The work describes the virtualization of a typical production process, the digital twin in the scope of Industry 4.0, involving different devices, such as robotic manipulators, conveyors, automatic warehouses and vision systems. It includes both legacy and recent equipment, with different characteristics and communication capabilities, ranging from RS232 serial communication to TCP/IP-based communication, or even I/O-based interaction for devices with no communication capabilities. The developed approach aims at industrial implementations, while allowing for educational purposes. For a standardized approach, the OPC UA protocol is used for high-level communication between the various systems. Several results are described showing the success of the methodology and application.

15:10-16:10 Session 13: Navigation
15:10
Development of a Navigator with Artificial Intelligence Applied to Smoothing and Execution of Paths in a Dynamic Two-Dimensional Environment

ABSTRACT. This paper shows the design and development of a navigation system that allows the execution of smoothed paths in a dynamic two-dimensional environment, which was applied to a quadruped robot.

In the proposal of this work, it is worth mentioning the use of the Robot Operating System (ROS) for the development and implementation of the navigator in a Raspberry Pi, Bezier Curve, used to smoothing the robot path, Video Tracking technique, that allows the identification of the movement of dynamic objects in the environment, and also the use of the Multilayer Perceptrons Neural Network in order to make the robot intelligent enough to get away from objects without leaving its route.

The proposed system allows a robot to move from a starting point to an end of a 2D environment without colliding with a moving object also inserted into the environment. This was designed, developed and tested in State University of Feira de Santana where we have a functional prototype. Although this prototype is running on a quadruped robot, it can be easily imported into other types of robots.

15:25
Monocular Visual Odometry Benchmarking and Turn Performance Optimization
SPEAKER: André Aguiar

ABSTRACT. Developing ground robots for crop monitoring and harvesting in steep slope vineyards is a complex challenge due to two main reasons: harsh condition of the terrain and unstable localization accuracy obtained with Global Navigation Satellite System. In this context, a reliable localization system requires an accurate and redundant information to Global Navigation Satellite System and wheel odometry based system. To pursue this goal we benchmark 3 well known Visual Odometry methods with 2 datasets. Two of these are feature-based Visual Odometry algorithms: Libviso2 and SVO 2.0. The third is an appearance- based Visual Odometry algorithm called DSO. In monocular Visual Odometry, two main problems appear: pure rotations and scale estimation. In this paper, we focus on the first issue. To do so, we propose a Kalman Filter to fuse a single gyroscope with the output pose of monocular Visual Odometry, while estimating gyroscope’s bias continuously. In this approach we propose a non-linear noise variation that ensures that bias estimation is not affected by Visual Odometry resultant rotations. We compare and discuss the three unchanged methods and the three methods with the proposed additional Kalman Filter. For tests, two public datasets are used: the Kitti dataset and another built in-house. Results show that our additional Kalman Filter highly improves Visual Odometry performance in rotation movements.

15:40
Q-Learning for Autonomous Mobile Robot Obstacle Avoidance

ABSTRACT. An approach to the problem of autonomous mobile robot obstacle avoidance using Reinforcement Learning, more precisely Q-Learning, is presented in this paper. Reinforcement Learning in Robotics has been a challenging topic for the past few years. The ability to equip a robot with a powerful enough tool to allow an autonomous discovery of an optimal behavior through trial-and-error interactions with its environment has been a reason for innumerous deep research projects. In this paper, two different Q-Learning approaches are presented as well as an extensive hyperparameter study. These algorithms were developed for a simplistically simulated Bot’n Roll ONE A (Fig. 1). The simulated robot communicates with the control script via ROS. The robot must surpass three levels of iterative complexity mazes similar to the ones presented on RoboParty [1] educational event challenge. For both algorithms, an extensive hyperparameter search was taken into account by testing hundreds of simulations with different parameters. Both Q-Learning solutions develop different strategies trying to solve the three labyrinths, enhancing its learning ability as well as discovering different approaches to certain situations, and finishing the task in complex environments

15:55
Environment-Aware Locomotion Mode Transition Prediction System

ABSTRACT. Current research suggests the emergent need to predict locomotion mode (LM) transitions to allow a natural and smooth response of lower limb active assistive devices such as prostheses and orthosis for daily life locomotion assistance. The purpose of this work is to develop an automatic, user-independent system using an environment-aware strategy to predict LM transitions. We applied an infrared laser system to measure the distance between the user and the terrain ahead. A three-layer decision tree with heuristic decision rules only dependent on laser features was implemented to predict LM transitions. The prediction system was validated with 10 healthy subjects that performed 8 LM transitions in different terrains (level-terrain, stairs, and ramps). The results showed a prediction accuracy above 80% for all LM transitions, achieving 100% prediction success for transitions ramp/stair descend to level-terrain. All the LM transitions were predicted with high prediction time (> 0.73 seconds) which empowers its integration on assistive devices control strategies. The prediction system accurately and time-effectively predicts 8 different LM transitions only using the laser sensor. It approached indoor and outdoor terrains, relevant for daily-life locomotion applications, and was more polyvalent and effective than previous environment-aware systems.

16:30-17:30 Session 14: Prosthetics and Limbs, Modeling & Control
16:30
Learning low level skills from scratch for humanoid robot soccer using deep reinforcement learning

ABSTRACT. Reinforcement learning algorithms are now more appealing than ever. Recent approaches bring power and tuning simplicity to the everyday work machine. The possibilities are endless, and the idea of automating learning without domain knowledge is quite tempting for many researchers. However, in competitive environments such as the RoboCup 3D Soccer Simulation League, there is a lot to be done regarding human-like behaviors. Current teams use many mechanical movements to perform basic skills, such as running and dribbling the ball. This paper aims to use the PPO algorithm to optimize those skills, achieving natural gaits without sacrificing performance. We use Simspark to simulate a NAO humanoid robot, using visual and body sensors to control its actuators. Based on our results, we propose an indirect control approach and detailed parameter setups to obtain natural running and dribbling behaviors. The obtained performance is in some cases comparable or better than the top RoboCup teams. However, some skills are not ready to be applied in competitive environments yet, due to instability. This work contributes towards the improvement of RoboCup and some related technical challenges.

16:45
A Model-Based Biped Walking Controller Based on Divergent Component of Motion

ABSTRACT. Biped robots have high degrees of freedom and they are naturally unstable, hence design and develop a reliable walking controller is a complex subject which is one of the interesting topics in the robotic community. In this paper, we proposed a model-based walking controller which is able to negate the effect of external impacts not only by applying compensating torques but also by adjusting the landing location of swing leg. This controller composed of two levels of control which takes into account the stable and unstable parts of COM dynamics. In our proposed controller, the overall dynamics of a humanoid robot is approximated using an enhanced version of Linear Inverted Pendulum Plus Flywheel Model (ELIPPFM) and according to this dynamics model, an optimal controller is designed to track the reference trajectories. Moreover, Divergent Component of Motion (DCM) is used to define when and where a robot should take a step to prevent falling. The proposed system has been successfully tested by performing several simulations using MATLAB. The results showed that the proposed system is capable of controlling the balance of the simulated robot in presence of severe disturbances.

17:00
Development of a simulated transtibial amputee model

ABSTRACT. Currently, there are more than 30 million amputees in the world and each year thousands of people suffer from amputation and, therefore, the development of lower limb prostheses is crucial to improve the quality of millions of people's lives by restoring their mobility. This contribution proposes a simulated amputee model capable of reproducing a transtibial amputee subject wearing a passive prosthesis. The passive prosthesis behavior is simulated using a spring-damper system between shin and foot. This contribution provides a tool capable of reproducing an amputee subject wearing a passive prosthesis, as well as an adaptive framework where researchers can deploy controllers in the simulated transtibial prosthesis, transforming it in a powered transtibial prostheses. Results show that the amputee model is good for the simulation of transtibial amputees wearing a passive device or an active transtibial prosthesis.

17:15
A Review on Commercially Available Anthropomorphic Myoelectric Prosthetic Hands, Pattern-Recognition-Based Microcontrollers and sEMG Sensors used for Prosthetic Control

ABSTRACT. It has been reported that over 3 million individuals live with upper limb amputation worldwide. Losing a hand drastically reduces the individual’s quality of life. Fortunately, there are several prosthetic solutions available in the market that try to restore some of the missing hand’s functions and characteristics. This paper presents a review on three of the main components of a typical transradial myoelectric prosthesis that can be found in the market. The goal was to provide the reader with an overview of commercially available anthropomorphic myoelectric prosthetic hands with high degrees of freedom, pattern-recognition-based microcontrollers and sEMG sensors used for prosthetic control.

17:40-18:30 Session 15: keynote 3

Robotic Dynamic Manipulation

Bruno Siciliano