Robotic Intelligence

The following sections are included:

• Foreword
• CONTENTS

In this context, a model is an algorithm or a procedure that applies to data resulting in a functional relation τ between “input space” $\mathcal{X}$ and “output space” $\mathcal{Y}$. In this short paper, we will delineate objective criteria which help to disambiguate and rate models’ credibility. We will define pertinent concepts and will voice an opinion on the matter of good versus bad versus so–so models.

Goodness-of-fit is used for the evaluation a model. They are commonly used to compare among competing models. The material is mostly classic. For more on the subject the reader is referred to the References including the two revised volumes Bickel and Docksum (2016).

It is not unusual that efforts to validate a statistical model exceed those used to build the model. Multiple techniques are used to validate, compare and contrast among competing statistical models: Some are concerned with a model’s ability to predict new data while others are concerned with model descriptiveness of the data. Without claiming to provide a comprehensive view of the landscape, in this paper we will touch on both aspects of model validation. There is much more to the subject and the reader is referred to any of the many classical statistical texts including the revised two volumes of Bickel and Docksum (2016), the one by Hastie, Tibshirani, and Friedman [The Elements of Statistical Learning: Data Mining, Inference, and Predication, 2nd edn. (Springer, 2009)], and several others listed in the bibliography.

Graph is a widely used scheme for representing complex datasets in terms of graphical illustration comprised of nodes and edges. Diffusion is a paradigm of propagating or transmitting substances or knowledge pieces from nodes to nodes. Diffusion analysis takes a slightly different approach from the reachability-based graph analysis; it takes the phenomenon as diffusion problems. In this paper, we present a technical survey on literatures of diffusion analysis.

Online Analytical Processing (OLAP) is an effective approach to analyzing various complex business problems, and graph is considered as a common scheme to represent the business datasets. Network analysis is a broad analytics scheme for exploring the connectivity and deriving useful analytics results. However, network analysis for graph-based OLAP presents a set of more specific analytics methods by utilizing graph model, network property, and OLAP principles. In this paper, we present a comprehensive survey on network analysis conducted on graph model for the purpose of OLAP, and we summarize the current research focus, paradigms, and the future needs on the target technology.

Microblogging platforms like Twitter, in the recent years, have become one of the important sources of information for a wide spectrum of users. As a result, these platforms have become great resources to provide support for emergency management. During any crisis, it is necessary to sieve through a huge amount of social media texts within a short span of time to extract meaningful information from them. Extraction of emergency-specific information, such as topic keywords or landmarks or geo-locations of sites, from these texts plays a significant role in building an application for emergency management. This paper thus highlights different aspects of automatic analysis of tweets to help in developing such an application. Hence, it focuses on: (1) identification of crisis-related tweets using machine learning, (2) exploration of topic model implementations and looking at its effectiveness on short messages (as short as 140 characters); and performing an exploratory data analysis on short texts related to crises collected from Twitter, and looking at different visualizations to understand the commonality and differences between topics and different crisis-related data, and (3) providing a proof of concept for identifying and retrieving different geo-locations from tweets and extracting the GPS coordinates from this data to approximately plot them in a map.

Academic institutions often assess the efficacy of courses by surveying students. These surveys are critical in structuring course content and evaluating instruction. Given the critical function of surveys for academic institutions, it is essential that surveys obtain data which is precise and accurate. Currently most institutions construct surveys employing the Likert scale: questions that require students to map their opinion on precise topics to a discrete, quantitative domain. These surveys uniformly weight response data, irrespective of student interest. We argue greater accuracy may be obtained by building Student-Directed Discussion Surveys (SDDSs) — surveys with several open-ended, student-directed questions, requiring free text responses. SDDSs retain precision by employing several Natural Language Processing (NLP) techniques including word frequency and sentiment analysis. We use SDDSs to improve course content and evaluate survey accuracy by comparing the results of an SDDS to a Likert-scaled survey administered to an overlapping population. We find that the results of these two survey techniques diverge when topics become increasingly significant to respondents. These results, in addition to the documented issues with Likert-scaled surveys, lead to the conclusion that SDDSs may provide more informative and insightful results.

There have been an enormous number of publications on cancer research. These unstructured cancer-related articles are of great value for cancer diagnostics, treatment, and prevention. The aim of this study is to introduce a recommendation system. It combines text mining (LDA) and semantic computing (GloVe) to understand the meaning of user needs and to increase the recommendation accuracy.

Although mereotopological relationship theories and their qualification problems have been extensively studied in ℝ², the qualification of mereotopological relations in ℝ³ remains challenging. This is due to the limited availability of topological operators and high costs of boundary intersection tests. In this paper, a novel qualification technique for mereotopological relations in ℝ³ is presented. Our technique rapidly computes RCC-8 base relations using precomputed signed distance fields, and makes no assumptions with regards to complexity or representation method of the spatial entities under consideration.

In this paper, a learning from demonstration (LFD) approach is used to design an autonomous meal-assistance agent. The feeding task is modeled as a mixture of Gaussian distributions. Using the data collected via kinesthetic teaching, the parameters of the Gaussian mixture model (GMM) are learnt using Gaussian mixture regression (GMR) and expectation maximization (EM) algorithm. Reproduction of feeding trajectories for different environments is obtained by solving a constrained optimization problem. In this method we show that obstacles can be avoided by robot’s end-effector by adding a set of extra constraints to the optimization problem. Finally, the performance of the designed meal assistant is evaluated in two feeding scenario experiments: one considering obstacles in the path between the bowl and the mouth and the other without.

The problem of cooperative obstacle avoidance by a group of unmanned ground vehicles (UGVs) and unmanned air vehicles (UAVs) during a typical search mission is addressed in this paper. The group of UAVs and UGVs are performing a search operation in the designated area. All the UAVs and UGVs are equipped with a vision sensor/LIDAR to identify the possible obstacles in the search space. Due to their operation on the ground, UGVs are more likely to encounter obstacles. The obstacle avoidance for UGV under the event of sensor failure is done with environment information from the nearest UAV. The UAV plans its trajectory according to the UGV’s expected future trajectory, leading towards the base station. The UGV replans its trajectory to avoid obstacles after obtaining the information from its nearest UAV. A simulation study is performed with 10 UAVs and five UGVs performing a search mission in 1 km × 1 km area. The proposed obstacle avoidance method is experimentally validated in the outdoor environment with an autonomous UAV equipped with a camera and an autonomous UGV navigating based on GPS localization and environment information from the UAV.

Robots are given more and more challenging tasks in domains such as transport and delivery, farming or health. Software is a key component for robots, and Robot Operating System (ROS) is a popular open-source middleware for writing robotics applications. Code quality matters a lot because a poorly written software is much more likely to contain bugs and will be harder to maintain over time. Within a code base, finding faulty patterns takes a lot of time and money. We propose a framework to search automatically user-provided faulty code patterns. This framework is based on FO⁺⁺, a temporal extension of first-order logic, and Pangolin, a verification engine for C++ programs. We formalized with FO⁺⁺ five faulty patterns related to ROS and embedded systems. We analyzed with Pangolin 25 ROS packages looking for occurrences of these patterns and found a total of 218 defects. To prevent the faulty patterns from arising in new ROS packages, we propose a design pattern, and we show how Pangolin can be used to enforce it.

Tasks that change the physical state of a robot and its environment take a considerable amount of time to execute. However, many robot applications spend the execution time waiting, although the following tasks might require time to prepare. This paper proposes to amend robot tasks with a semantic description of their expected outcomes, which allows planning and preparing successive tasks based on this information. The suggested approach allows sequential and parallel compositions of tasks, as well as reactive behavior modeled as state machines. The paper describes the means of modeling and executing these tasks, details different possibilities of planning in state-machine tasks and evaluates the benefits achievable using the approach on two example scenarios.

Word-play is as powerful learning and motivation tool often used by educators for teaching the ability of reading, which is a complex activity. In this paper, we introduce a system that exploits a Pepper humanoid robot acting as a playfellow in a word-play game based on portmanteau words. The robot shows the ability to play with children using a conversation engine, a portmanteau creation engine, and a definition engine. In this manner, Pepper can integrate itself within a group of kids, and it can support a teacher in her activities. The humanoid can be involved in a word-based round-game in which it can play the role of either answerer or generator of new words.

A key requirement in the development of intelligent and driverless vehicles is steering angle computation for efficient navigation. This paper presents a novel method for computing steering angle for driverless vehicles using computer vision-based techniques of relatively lower computing cost. The proposed system consists of four major stages. The first stage includes dynamic road region extraction using Gaussian mixture model and expectation maximization algorithm. The second stage is to compute the steering angle based on the extracted road region. Subsequently, Kalman filtering technique is used to cancel spurious angle transition noises. In addition, future steering angle is estimated which in turn gives informative feedback for smooth navigation of the vehicle. The proposed algorithm was tested both on a simulator and real-time images and was found to give a good estimation of actual steering angle required for navigation. Further, it was also observed that this works in different lighting conditions as well as for both structured and unstructured road scenarios.

This paper focuses on enhancing the mission duration by deploying secondary agents to coordinate with the primary agents to accomplish the mission with a minimal interruption. The interruption considered here is due to limited fuel carrying capability of primary agents. In this study, primary and secondary agents refer to unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs), respectively. Conventionally, UAVs are refueled with the fixed main charging stations which lead to interruption during the ongoing mission. In this work, we propose two-stage density estimation approach for efficiently distributing the swarm of UGVs to act as mobile refueling stations for UAVs. In the first stage, the optimal number of UGVs and their initial placement are computed. In the final stage, the UGVs minimize the average distance for the nearest UAVs to refuel. The performance of the proposed method is compared with the state of the art. The numerical simulation shows a better performance with the distributed UGVs than the state of the art.

In the paper, we describe a trajectory planning problem for a six-DoF robotic manipulator arm that carries an ultra-wideband (UWB) radar sensor with synthetic aperture (SAR). The resolution depends on the trajectory and velocity profile of the sensor head. The constraints can be modeled as an optimization problem to obtain a feasible, collision-free target trajectory of the end-effector of the manipulator arm in Cartesian coordinates that minimizes observation time. For 3D reconstruction, the target is observed in multiple height slices. For through-the-wall radar the sensor can be operated in sliding mode for scanning larger areas. For IED inspection the spotlight mode is preferred, constantly pointing the antennas towards the target to obtain maximum azimuth resolution. UWB sensors typically use a wide spectrum shared by other RF communication systems. This may become a limiting factor on system sensitivity and severely degrade the image quality. Cognitive radars can adapt dynamically their bandwidth, frequency and other transmit parameters to the radio frequency environment to avoid interference with primary users.

The rapid advances in robotics have recently led to the developments of a wide range of robotic platforms that exhibit significant differences at the hardware components level. Consequently, this poses a significant challenge to robot software developers since they have to know how every hardware device in the robot works to ensure their software’s compatibility when transferring/reusing their code on different robots. In this paper we present a new Robot Hardware Abstraction Layer (R-HAL) that permits to seamlessly program and control any robotic platform powered by the XBot control software framework. The implementation details of the R-HAL are introduced. The R-HAL is extensively validated through simulation trials and experiments with a wide range of dissimilar robotic platforms, among them the COMAN and WALK-MAN humanoids, the KUKA LWR and the CENTAURO upper body. The results attained demonstrate in practice the gained benefits in terms of code compatibility, reuse and portability, and finally unified application programming even for robots with significantly diverse hardware.

The paper illustrates a cognitive architecture for computational creativity based on the Psi model and the mechanisms inspired by dual-process theories of reasoning and rationality. In particular, three applications of computational creativity will be summed up: (a) a robot capable of executing creative paintings through a multilayer mechanism that implements an associative memory and is capable to properly mix elements belonging to different domains; (b) a robot aimed at producing a collage formed by a mix of photo-montage and digital collage: the artwork is created after a visual and verbal interaction with a human user; and (c) a humanoid robot capable of improvising a dancing choreography in real-time according to the listened music. First results in computational creativity show a set of potentialities to be explored that can shed light on human and artificial creativities and artificial intelligent systems.

The following sections are included:

• Subject Index
• Author Index