Eurico Pedrosa

Online SLAM Based on a Fast Scan-Matching Algorithm

This paper presents a scan-matching approach for online simultaneous localization and mapping. This approach combines a fast and efficient scan-matching algorithm for localization with dynamic and approximate likelihood fields to incrementally build a map. The achievable results of the approach are evaluated using an objective benchmark designed to compare SLAM solutions that use different methods. The result is a fast online SLAM approach suitable for real-time operations.

A Skill-Based Architecture for Pick and Place Manipulation Tasks

Robots can play a significant role in product customization but they should leave a repetitive, low intelligence paradigm and be able to operate in unstructured environments and take decisions during the execution of the task. The EuRoC research project addresses this issue by posing as a competition to motivate researchers to present their solution to the problem. The first stage is a simulation competition where Pick & Place type of tasks are the goal and planning, perception and manipulation are the problems. This paper presents a skill-based architecture that enables a simulated moving manipulator to solve these tasks. The heuristics that were used to solve specific tasks are also presented. Using computer vision methods and the definition of a set of manipulation skills, an intelligent agent is able to solve them autonomously. The work developed in this project was used in the simulation competition of EuRoC project by team IRIS and enabled them to reach the \(5^{\mathrm {th}}\) rank.

Rich and robust human-robot interaction on gesture recognition for assembly tasks

The adoption of robotics technology has the potential to advance quality, efficiency and safety for manufacturing enterprises, in particular small and medium-sized enterprises. This paper presents a human-robot interaction (HRI) system that enables a robot to receive commands, provide information to a human teammate and ask them a favor. In order to build a robust HRI system based on gesture recognition, three key issues are addressed: richness, multiple feature fusion and failure verification. The developed system has been tested and validated in a realistic lab with a real mobile manipulator and a human teammate to solve a puzzle game.

Skill-based anytime agent architecture for logistics and manipulation tasks: EuRoC Challenge 2, Stage II - Realistic Labs: Benchmarking

Nowadays, the increase of robotic technology application to industry scenarios is notorious. Proposals for new effective solutions are in continuous development once industry needs a constantly improvement in time as well as in production quality and efficiency. The EuRoC research project proposes a scientific competition in which research and industry manufacturers joint teams are encouraged to develop and test solutions that can solve several issues as well as be useful in manufacturing improvement. This paper presents the TIMAIRIS architecture and approach used in the Challenge 2 - Stage II - Benchmarking phase, namely regarding the perception, manipulation and planning strategy that was applied to achieve the tasks objectives. The used approach proved to be quite robust and efficient, which allowed us to rank first in the Benchmarking phase.

Improving and Benchmarking Motion Planning for a Mobile Manipulator Operating in Unstructured Environments.

This paper presents the use, adaptation and benchmarking of motion planning tools that will be integrated with the KUKA KMR iiwa mobile robot. The motion planning tools are integrated in the robotic agent presented in [1]. The adaptation consists on algorithms developed to increase the robustness and the efficiency to solve the motion planning problems. These algorithms combine existing motion planners with a trajectory filter developed in this work. Finally, the benchmarking of different motion planners is presented. Three motion planning tasks with a growing level of complexity are taken in consideration for the tests in a simulation environment. The motion planners that provided the best results were RRTConnect for the two less complex tasks and PRM* for the most difficult task.

A Scan Matching Approach to SLAM with a Dynamic Likelihood Field

This paper presents a fast scan matching approach to online SLAM supported by a dynamic likelihood field. The dynamic likelihood field plays a central role in the approach, as it avoids the necessity to establish direct correspondences, it is the connection link between scan matching and the online SLAM and it has a low computational complexity. Scan matching is formulated as a non-linear least squares problem and solved by the Gauss-Newton method. Furthermore, to reduce the influences of outliers during optimization, a loss function is introduced. The proposed solution was evaluated using an objective benchmark designed to compare SLAM solutions and its execution times were also analyzed. It shows to be a fast and accurate online SLAM approach, suitable for real-time operation.

Human-Robot Collaboration and Safety Management for Logistics and Manipulation Tasks

To realize human-robot collaboration in manufacturing, industrial robots need to share an environment with humans and to work hand in hand. This introduces safety concerns but also provides the opportunity to take advantage of human-robot interactions to control the robot. The main objective of this work is to provide HRI without compromising safety issues in a realistic industrial context. In the paper, a region-based filtering and reasoning method for safety has been developed and integrated into a human-robot collaboration system. The proposed method has been successfully demonstrated keeping safety during the showcase evaluation of the European robotics challenges with a real mobile manipulator.

Neural regularization jointly involving neurons and connections for robust image classification

This paper presents an integrated neural regularization method in fully-connected neural networks that jointly combines the cutting edge of regularization techniques; Dropout [1] and DropConnect [2]. With a small number of data set, trained feed-forward networks tend to show poor prediction performance on test data which has never been introduced while training. In order to reduce the overfitting, regularization methods commonly use only a sparse subset of their inputs. While a fully-connected layer with Dropout takes account of a randomly selected subset of hidden neurons with some probability, a layer with DropConnect only keeps a randomly selected subset of connections between neurons. It has been reported that their performances are dependent on domains. Image classification results show that the integrated method provides more degrees of freedom to achieve robust image recognition in the test phase. The experimental analyses on CIFAR-10 and one-hand gesture dataset show that the method provides the opportunity to improve classification performance.

Efficient localization based on scan matching with a continuous likelihood field

This paper presents a fast scan matching approach to mobile robot localization supported by a continuous likelihood field. The likelihood field plays a central role in the approach, as it avoids the necessity to establish direct correspondences; it is the connection link between scan matching and robotic localization, and it provides a reduced computational complexity. Scan matching is formulated as a non-linear least squares problem and solved by the Gauss-Newton and Levenberg-Marquardt methods. Furthermore, to reduce the influences of outliers during optimization, a loss function is introduced. The proposed solution was evaluated using a publicly available dataset and compared with AMCL, a state-of-the-art localization algorithm. Our proposal shows to be a fast and accurate localization algorithm suitable for any type of operation.

A Non-Linear Least Squares Approach to SLAM using a Dynamic Likelihood Field

This paper presents a fast scan matching approach to online SLAM supported by a dynamic likelihood field. The dynamic likelihood field plays a central role in the approach: it avoids the necessity to establish direct correspondences; it is the connection link between scan matching and the online SLAM; and it has a low computational complexity. Scan matching is formulated as a non-linear least squares problem that allows us to solve it using Gauss-Newton or Levenberg-Marquardt methods. Furthermore, to reduce the influence of outliers during optimization, a loss function is introduced. The proposed solution was evaluated using an objective benchmark designed to compare different SLAM solutions. Additionally, the execution times of our proposal were also analyzed. The obtained results show that the proposed approach provides a fast and accurate online SLAM, suitable for real-time operation.