Chapter 2. Symbolic Reasoning in Latent Space: Classical Planning as an Example

Symbolic systems require hand-coded symbolic representation as input, resulting in a knowledge acquisition bottleneck. Meanwhile, although deep learning has achieved significant success in many fields, the knowledge is encoded in a subsymbolic representation which is incompatible with symbolic systems. To address the gap between the two fields, one has to solve Symbol Grounding problem: The question of how a machine can generate symbols automatically. We discuss our recent work called Latplan, an unsupervised architecture combining deep learning and classical planning. Given only an unlabeled set of image pairs showing a subset of transitions allowed in the environment (training inputs), Latplan learns a complete propositional PDDL action model of the environment. Later, when a pair of images representing the initial and the goal states (planning inputs) is given, Latplan finds a plan to the goal state in a symbolic latent space and returns a visualized plan execution. We discuss several key ideas that made Latplan possible which would hopefully extend to many other symbolic paradigms outside classical planning.

Using Platform Models for a Guided Explanatory Diagnosis Generation for Mobile Robots

Plan execution on a mobile robot is inherently error-prone, as the robot needs to act in a physical world which can never be completely controlled by the robot. If an error occurs during execution, the true world state is unknown, as a failure may have unobservable consequences. One approach to deal with such failures is diagnosis, where the true world state is determined by identifying a set of faults based on sensed observations. In this paper, we present a novel approach to explanatory diagnosis, based on the assumption that most failures occur due to some robot hardware failure. We model the robot platform components with state machines and formulate action variants for the robots' actions, modelling different fault modes. We apply diagnosis as planning with a top-k planning approach to determine possible diagnosis candidates and then use active diagnosis to find out which of those candidates is the true diagnosis. Finally, based on the platform model, we recover from the occurred failure such that the robot can continue to operate. We evaluate our approach in a logistics robots scenario by comparing it to having no diagnosis and diagnosis without platform models, showing a significant improvement to both alternatives.

Development of a Checklist and Analysis on the Case of Design Thinking Activity in University

The purposes of this study were to develop a checklist, which guides the process of design thinking for the students, and to analyze the case of A university, so as to facilitate the non-subject educational activities in university. From the results of the literature review, we derived the preliminary items included in the checklist. After verifying the content validity, a Delphi survey by 15 experts was conducted. Corresponding to the five steps of design thinking, a total of 22 items, including 3 items for ‘empathize’, 6 items for ‘define’, 5 items for ‘idea’, 3 items for ‘prototype’, and 5 items for ‘test’ were finally selected for the students’ checklist.With regard to their problem solving abilities, the design thinking based program might be beneficial for enhancing students’ competence when it comes to problem clarification and plan execution. Furthermore, the developed design thinking checklists, along with the presented case of the design thinking program, are expected to encourage students to more actively apply the design thinking process to their non-subject educational activities.

Goal Lifecycle Networks For Robotics

A Goal Lifecycle Network (GLN) is a conceptual process model that captures the progression of goals from their formulation to their completion, including planning and execution concerns. GLNs synthesize the literature on hierarchical goal networks, goal lifecycles, and plan execution. We formalize GLNs based on a state-variable representation, extend GLNs with an execution lifecycle, describe a partial reference implementation of GLNs, and show how the temporal PDDL language can be translated into GLNs for dispatchable execution. We integrate GLNs in three proof-of-concept robotics demonstrations: (1) a two-armed robot sorting items into baskets; (2) a multi-vehicle quad-rotor team surveying a region; and (3) centralized planning for a simulated disaster relief based on the Robocup Rescue League. The theory, implementation, and demonstrations highlight that GLNs are effective for goal management in the robotics systems we study.

Internet-of-Things Skills Among the General Population: Task-Based Performance Test Using Activity Trackers

Background The health internet-of-things (IoT) can potentially provide insights into the present health condition, potential pitfalls, and support of a healthier lifestyle. However, to enjoy these benefits, people need skills to use the IoT. These IoT skills are expected to differ across the general population, thereby causing a new digital divide. Objective This study aims to assess whether a sample of the general Dutch population can use health IoT by focusing on data and strategic IoT skills. Furthermore, we determine the role of gender, age, and education, and traditional internet skills. Methods From April 1, 2019, to December 12, 2019, 100 individuals participated in this study. Participants were recruited via digital flyers and door-to-door canvassing. A selective quota sample was divided into equal subsamples of gender, age, and education. Additional inclusion criteria were smartphone possession and no previous experience of using activity trackers. This study was conducted in 3 waves over a period of 2 weeks. In wave 1, a questionnaire was administered to measure the operational, mobile, and information internet skills of the participants, and the participants were introduced to the activity tracker. After 1 week of getting acquainted with the activity tracker, a task-based performance test was conducted in wave 2 to measure the levels of data IoT skills and the strategic IoT skill component—action plan construction. A week after the participants were asked to use the activity tracker more deliberately, a performance test was then conducted in wave 3 to measure the level of the strategic IoT skill component—action plan execution. Results The participants successfully completed 54% (13.5/25) of the data IoT skill tasks. Regarding strategic IoT tasks, the completion rates were 56% (10.1/18) for action plan construction and 43% (3.9/9) for action plan execution. None of the participants were able to complete all the data IoT skill tasks, and none of the participants were able to complete all the strategic IoT skill tasks regarding action plan construction or its execution. Age and education were important determinants of the IoT skill levels of the participants, except for the ability to execute an action plan strategically. Furthermore, the level of information internet skills of the participants contributed to their level of data IoT skills. Conclusions This study found that data and strategic IoT skills of Dutch citizens are underdeveloped with regard to health purposes. In particular, those who could benefit the most from health IoT were those who had the most trouble using it, that is, the older and lower-educated individuals.

The four-phase team adaptation process: a first empirical investigation

Purpose Several theoretical models have been developed to describe the process of successful team adaptation. Testing the models through empirical research is lacking. This study aims to empirically examine the way teams adapt to unexpected or novel circumstances and investigate the four-phase team adaptation process (i.e. situation assessment → plan formulation → plan execution → team learning), as proposed by Rosen et al. (2011). Design/methodology/approach To test the positive relationship between the four team adaptation phases and their suggested sequence, a cross-sectional field study was conducted. Data were collected from 23 teams participating during an 8-week team project. Findings Results from random intercept models confirmed that the team adaptation process consisted of four phases that were positively related to each other. As expected, plan formulation mediated the positive relationship between situation assessment and plan execution. However, team learning was independently related to all three previous phases, and not only to situation assessment as theory suggests. Originality/value To the best of the authors’ knowledge, the present study is one of the first attempts to test the theoretical model of the team adaptation process presented by Rosen et al. (2011). Findings illustrated that the team adaptation process is not a simple four-phase sequence, but it constitutes four dynamic phases that are strongly interrelated to each other.

Internet-of-Things Skills Among the General Population: Task-Based Performance Test Using Activity Trackers (Preprint)

BACKGROUND The health internet-of-things (IoT) can potentially provide insights into the present health condition, potential pitfalls, and support of a healthier lifestyle. However, to enjoy these benefits, people need skills to use the IoT. These <i>IoT skills</i> are expected to differ across the general population, thereby causing a new digital divide. OBJECTIVE This study aims to assess whether a sample of the general Dutch population can use health IoT by focusing on data and strategic IoT skills. Furthermore, we determine the role of gender, age, and education, and <i>traditional</i> internet skills. METHODS From April 1, 2019, to December 12, 2019, 100 individuals participated in this study. Participants were recruited via digital flyers and door-to-door canvassing. A selective quota sample was divided into equal subsamples of gender, age, and education. Additional inclusion criteria were smartphone possession and no previous experience of using activity trackers. This study was conducted in 3 waves over a period of 2 weeks. In wave 1, a questionnaire was administered to measure the operational, mobile, and information internet skills of the participants, and the participants were introduced to the activity tracker. After 1 week of getting acquainted with the activity tracker, a task-based performance test was conducted in wave 2 to measure the levels of data IoT skills and the strategic IoT skill component—<i>action plan construction</i>. A week after the participants were asked to use the activity tracker more deliberately, a performance test was then conducted in wave 3 to measure the level of the strategic IoT skill component—<i>action plan execution</i>. RESULTS The participants successfully completed 54% (13.5/25) of the data IoT skill tasks. Regarding strategic IoT tasks, the completion rates were 56% (10.1/18) for action plan construction and 43% (3.9/9) for action plan execution. None of the participants were able to complete all the data IoT skill tasks, and none of the participants were able to complete all the strategic IoT skill tasks regarding action plan construction or its execution. Age and education were important determinants of the IoT skill levels of the participants, except for the ability to execute an action plan strategically. Furthermore, the level of information internet skills of the participants contributed to their level of data IoT skills. CONCLUSIONS This study found that data and strategic IoT skills of Dutch citizens are underdeveloped with regard to health purposes. In particular, those who could benefit the most from health IoT were those who had the most trouble using it, that is, the older and lower-educated individuals.