Search

As AI agents leave the lab and venture into the real world as autonomous vehicles, delivery robots, and cooking robots, it is increasingly necessary to design and comprehensively evaluate algorithms that tackle the ``open-world''. To this end, we introduce NovelGym, a flexible and adaptable ecosystem designed to simulate gridworld environments, serving as a robust platform for benchmarking reinforcement learning (RL) and hybrid planning and learning agents in open-world contexts. The modular architecture of NovelGym facilitates rapid creation and modification of task environments, including multi-agent scenarios, with multiple environment transformations, thus providing a dynamic testbed for researchers to develop open-world AI agents., Comment: Accepted at AAMAS-2024

Domain generalization aims to learn a model with good generalization ability, that is, the learned model should not only perform well on several seen domains but also on unseen domains with different data distributions. State-of-the-art domain generalization methods typically train a representation function followed by a classifier jointly to minimize both the classification risk and the domain discrepancy. However, when it comes to model selection, most of these methods rely on traditional validation routines that select models solely based on the lowest classification risk on the validation set. In this paper, we theoretically demonstrate a trade-off between minimizing classification risk and mitigating domain discrepancy, i.e., it is impossible to achieve the minimum of these two objectives simultaneously. Motivated by this theoretical result, we propose a novel model selection method suggesting that the validation process should account for both the classification risk and the domain discrepancy. We validate the effectiveness of the proposed method by numerical results on several domain generalization datasets., Comment: Accepted to ICASSP 2023

Learning to detect, characterize and accommodate novelties is a challenge that agents operating in open-world domains need to address to be able to guarantee satisfactory task performance. Certain novelties (e.g., changes in environment dynamics) can interfere with the performance or prevent agents from accomplishing task goals altogether. In this paper, we introduce general methods and architectural mechanisms for detecting and characterizing different types of novelties, and for building an appropriate adaptive model to accommodate them utilizing logical representations and reasoning methods. We demonstrate the effectiveness of the proposed methods in evaluations performed by a third party in the adversarial multi-agent board game Monopoly. The results show high novelty detection and accommodation rates across a variety of novelty types, including changes to the rules of the game, as well as changes to the agent's action capabilities.

Domain generalization (DG) is a branch of transfer learning that aims to train the learning models on several seen domains and subsequently apply these pre-trained models to other unseen (unknown but related) domains. To deal with challenging settings in DG where both data and label of the unseen domain are not available at training time, the most common approach is to design the classifiers based on the domain-invariant representation features, i.e., the latent representations that are unchanged and transferable between domains. Contrary to popular belief, we show that designing classifiers based on invariant representation features alone is necessary but insufficient in DG. Our analysis indicates the necessity of imposing a constraint on the reconstruction loss induced by representation functions to preserve most of the relevant information about the label in the latent space. More importantly, we point out the trade-off between minimizing the reconstruction loss and achieving domain alignment in DG. Our theoretical results motivate a new DG framework that jointly optimizes the reconstruction loss and the domain discrepancy. Both theoretical and numerical results are provided to justify our approach., Comment: 13 pages, 2 tables

In this paper, we propose a novel domain generalization (DG) framework based on a new upper bound to the risk on the unseen domain. Particularly, our framework proposes to jointly minimize both the covariate-shift as well as the concept-shift between the seen domains for a better performance on the unseen domain. While the proposed approach can be implemented via an arbitrary combination of covariate-alignment and concept-alignment modules, in this work we use well-established approaches for distributional alignment namely, Maximum Mean Discrepancy (MMD) and covariance Alignment (CORAL), and use an Invariant Risk Minimization (IRM)-based approach for concept alignment. Our numerical results show that the proposed methods perform as well as or better than the state-of-the-art for domain generalization on several data sets., Comment: 8 pages, 2 figures, and 1 table. This paper is accepted at 32nd IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2022)

We propose RAPid-Learn: Learning to Recover and Plan Again, a hybrid planning and learning method, to tackle the problem of adapting to sudden and unexpected changes in an agent's environment (i.e., novelties). RAPid-Learn is designed to formulate and solve modifications to a task's Markov Decision Process (MDPs) on-the-fly and is capable of exploiting domain knowledge to learn any new dynamics caused by the environmental changes. It is capable of exploiting the domain knowledge to learn action executors which can be further used to resolve execution impasses, leading to a successful plan execution. This novelty information is reflected in its updated domain model. We demonstrate its efficacy by introducing a wide variety of novelties in a gridworld environment inspired by Minecraft, and compare our algorithm with transfer learning baselines from the literature. Our method is (1) effective even in the presence of multiple novelties, (2) more sample efficient than transfer learning RL baselines, and (3) robust to incomplete model information, as opposed to pure symbolic planning approaches., Comment: Proceedings of the IEEE Conference on Development and Learning (ICDL 2022)

In order for artificial agents to successfully perform tasks in changing environments, they must be able to both detect and adapt to novelty. However, visual novelty detection research often only evaluates on repurposed datasets such as CIFAR-10 originally intended for object classification, where images focus on one distinct, well-centered object. New benchmarks are needed to represent the challenges of navigating the complex scenes of an open world. Our new NovelCraft dataset contains multimodal episodic data of the images and symbolic world-states seen by an agent completing a pogo stick assembly task within a modified Minecraft environment. In some episodes, we insert novel objects of varying size within the complex 3D scene that may impact gameplay. Our visual novelty detection benchmark finds that methods that rank best on popular area-under-the-curve metrics may be outperformed by simpler alternatives when controlling false positives matters most. Further multimodal novelty detection experiments suggest that methods that fuse both visual and symbolic information can improve time until detection as well as overall discrimination. Finally, our evaluation of recent generalized category discovery methods suggests that adapting to new imbalanced categories in complex scenes remains an exciting open problem., Comment: Published in Transactions on Machine Learning Research (03/2023)

As AI-enabled robots enter the realm of healthcare and caregiving, it is important to consider how they will address the dimensions of care and how they will interact not just with the direct receivers of assistance, but also with those who provide it (e.g., caregivers, healthcare providers etc.). Caregiving in its best form addresses challenges in a multitude of dimensions of a person's life: from physical, to social-emotional and sometimes even existential dimensions (such as issues surrounding life and death). In this study we use semi-structured qualitative interviews administered to healthcare professions with multidisciplinary backgrounds (physicians, public health professionals, social workers, and chaplains) to understand their expectations regarding the possible roles robots may play in the healthcare ecosystem in the future. We found that participants drew inspiration in their mental models of robots from both works of science fiction but also from existing commercial robots. Participants envisioned roles for robots in the full spectrum of care, from physical to social-emotional and even existential-spiritual dimensions, but also pointed out numerous limitations that robots have in being able to provide comprehensive humanistic care. While no dimension of care was deemed as exclusively the realm of humans, participants stressed the importance of caregiving humans as the primary providers of comprehensive care, with robots assisting with more narrowly focused tasks. Throughout the paper we point out the encouraging confluence of ideas between the expectations of healthcare providers and research trends in the human-robot interaction (HRI) literature.

Invariance-principle-based methods such as Invariant Risk Minimization (IRM), have recently emerged as promising approaches for Domain Generalization (DG). Despite promising theory, such approaches fail in common classification tasks due to the mixing of true invariant features and spurious invariant features. To address this, we propose a framework based on the conditional entropy minimization (CEM) principle to filter-out the spurious invariant features leading to a new algorithm with a better generalization capability. We show that our proposed approach is closely related to the well-known Information Bottleneck (IB) framework and prove that under certain assumptions, entropy minimization can exactly recover the true invariant features. Our approach provides competitive classification accuracy compared to recent theoretically-principled state-of-the-art alternatives across several DG datasets., Comment: 10 pages, this paper was accepted at 26th International Conference on Pattern Recognition (ICPR-2022)

Dialogue agents that interact with humans in situated environments need to manage referential ambiguity across multiple modalities and ask for help as needed. However, it is not clear what kinds of questions such agents should ask nor how the answers to such questions can be used to resolve ambiguity. To address this, we analyzed dialogue data from an interactive study in which participants controlled a virtual robot tasked with organizing a set of tools while engaging in dialogue with a live, remote experimenter. We discovered a number of novel results, including the distribution of question types used to resolve ambiguity and the influence of dialogue-level factors on the reference resolution process. Based on these empirical findings we: (1) developed a computational model for clarification requests using a decision network with an entropy-based utility assignment method that operates across modalities, (2) evaluated the model, showing that it outperforms a slot-filling baseline in environments of varying ambiguity, and (3) interpreted the results to offer insight into the ways that agents can ask questions to facilitate situated reference resolution., Comment: To be published in the proceedings of the 23rd ACM International Conference on Multimodal Interaction (ICMI) 2021

This paper advances the theory and practice of Domain Generalization (DG) in machine learning. We consider the typical DG setting where the hypothesis is composed of a representation mapping followed by a labeling function. Within this setting, the majority of popular DG methods aim to jointly learn the representation and the labeling functions by minimizing a well-known upper bound for the classification risk in the unseen domain. In practice, however, methods based on this theoretical upper bound ignore a term that cannot be directly optimized due to its dual dependence on both the representation mapping and the unknown optimal labeling function in the unseen domain. To bridge this gap between theory and practice, we introduce a new upper bound that is free of terms having such dual dependence, resulting in a fully optimizable risk upper bound for the unseen domain. Our derivation leverages classical and recent transport inequalities that link optimal transport metrics with information-theoretic measures. Compared to previous bounds, our bound introduces two new terms: (i) the Wasserstein-2 barycenter term that aligns distributions between domains, and (ii) the reconstruction loss term that assesses the quality of representation in reconstructing the original data. Based on this new upper bound, we propose a novel DG algorithm named Wasserstein Barycenter Auto-Encoder (WBAE) that simultaneously minimizes the classification loss, the barycenter loss, and the reconstruction loss. Numerical results demonstrate that the proposed method outperforms current state-of-the-art DG algorithms on several datasets., Comment: This article has been accepted for publication in IEEE Access

The game of monopoly is an adversarial multi-agent domain where there is no fixed goal other than to be the last player solvent, There are useful subgoals like monopolizing sets of properties, and developing them. There is also a lot of randomness from dice rolls, card-draws, and adversaries' strategies. This unpredictability is made worse when unknown novelties are added during gameplay. Given these challenges, Monopoly was one of the test beds chosen for the DARPA-SAILON program which aims to create agents that can detect and accommodate novelties. To handle the game complexities, we developed an agent that eschews complete plans, and adapts it's policy online as the game evolves. In the most recent independent evaluation in the SAILON program, our agent was the best performing agent on most measures. We herein present our approach and results.

Understanding the spread of false or dangerous beliefs through a population has never seemed so urgent. Network science researchers have often taken a page from epidemiologists, and modeled the spread of false beliefs as similar to how a disease spreads through a social network. However, absent from those disease-inspired models is an internal model of an individual's set of current beliefs, where cognitive science has increasingly documented how the interaction between mental models and incoming messages seems to be crucially important for their adoption or rejection. Some computational social science modelers analyze agent-based models where individuals do have simulated cognition, but they often lack the strengths of network science, namely in empirically-driven network structures. We introduce a cognitive cascade model that combines a network science belief cascade approach with an internal cognitive model of the individual agents as in opinion diffusion models as a public opinion diffusion (POD) model, adding media institutions as agents which begin opinion cascades. We conduct an analysis of the cognitive cascade model with our simple cognitive function across various graph topologies and institutional messaging patterns. We argue from our results that population-level aggregate outcomes of the model qualitatively match what has been reported in COVID-related public opinion polls, and that the model dynamics lend insights as to how to address the spread of problematic beliefs. The overall model sets up a framework with which social science misinformation researchers and computational opinion diffusion modelers can join forces to understand, and hopefully learn how to best counter, the spread of disinformation and "alternative facts."

Intelligent agents that are confronted with novel concepts in situated environments will need to ask their human teammates questions to learn about the physical world. To better understand this problem, we need data about asking questions in situated task-based interactions. To this end, we present the Human-Robot Dialogue Learning (HuRDL) Corpus - a novel dialogue corpus collected in an online interactive virtual environment in which human participants play the role of a robot performing a collaborative tool-organization task. We describe the corpus data and a corresponding annotation scheme to offer insight into the form and content of questions that humans ask to facilitate learning in a situated environment. We provide the corpus as an empirically-grounded resource for improving question generation in situated intelligent agents., Comment: Corpus available at https://github.com/USArmyResearchLab/ARL-HuRDL . To appear in proceedings of SIGDial 2021

Trust in human-robot interactions (HRI) is measured in two main ways: through subjective questionnaires and through behavioral tasks. To optimize measurements of trust through questionnaires, the field of HRI faces two challenges: the development of standardized measures that apply to a variety of robots with different capabilities, and the exploration of social and relational dimensions of trust in robots (e.g., benevolence). In this paper we look at how different trust questionnaires fare given these challenges that pull in different directions (being general vs. being exploratory) by studying whether people think the items in these questionnaires are applicable to different kinds of robots and interactions. In Study 1 we show that after being presented with a robot (non-humanoid) and an interaction scenario (fire evacuation), participants rated multiple questionnaire items such as "This robot is principled" as "Non-applicable to robots in general" or "Non-applicable to this robot". In Study 2 we show that the frequency of these ratings change (indeed, even for items rated as N/A to robots in general) when a new scenario is presented (game playing with a humanoid robot). Finally, while overall trust scores remained robust to N/A ratings, our results revealed potential fallacies in the way these scores are commonly interpreted. We conclude with recommendations for the development, use and results-reporting of trust questionnaires for future studies, as well as theoretical implications for the field of HRI., Comment: 9 pages

Regular irradiation of indoor environments with ultraviolet C (UVC) light has become a regular task for many indoor settings as a result of COVID-19, but current robotic systems attempting to automate it suffer from high costs and inefficient irradiation. In this paper, we propose a purpose-made inexpensive robotic platform with off-the-shelf components and standard navigation software that, with a novel algorithm for finding optimal irradiation locations, addresses both shortcomings to offer affordable and efficient solutions for UVC irradiation. We demonstrate in simulations the efficacy of the algorithm and show a prototypical run of the autonomous integrated robotic system in an indoor environment. In our sample instances, our proposed algorithm reduces the time needed by roughly 30\% while it increases the coverage by a factor of 35\% (when compared to the best possible placement of a static light)., Comment: Preliminary version of this paper will be published in the ICRA 2021 conference

Symbolic planning models allow decision-making agents to sequence actions in arbitrary ways to achieve a variety of goals in dynamic domains. However, they are typically handcrafted and tend to require precise formulations that are not robust to human error. Reinforcement learning (RL) approaches do not require such models, and instead learn domain dynamics by exploring the environment and collecting rewards. However, RL approaches tend to require millions of episodes of experience and often learn policies that are not easily transferable to other tasks. In this paper, we address one aspect of the open problem of integrating these approaches: how can decision-making agents resolve discrepancies in their symbolic planning models while attempting to accomplish goals? We propose an integrated framework named SPOTTER that uses RL to augment and support ("spot") a planning agent by discovering new operators needed by the agent to accomplish goals that are initially unreachable for the agent. SPOTTER outperforms pure-RL approaches while also discovering transferable symbolic knowledge and does not require supervision, successful plan traces or any a priori knowledge about the missing planning operator., Comment: Accepted to AAMAS 2021

Autonomous robots with sophisticated capabilities can make it difficult for human instructors to assess its capabilities and proficiencies. Therefore, it is important future robots have the ability to: introspect on their capabilities and assess their task performance. Introspection allows the robot to determine what it can accomplish and self-assessment allows the robot estimate the likelihood it will accomplish at given task. We introduce a general framework for introspection and self-assessment that enables robots to have task and performance-based dialogues before, during, and after a mission. We then realize aspects of the framework in the cognitive robotic DIARC architecture, and finally show a proof-of-concept demonstration on a Nao robot showing its self-assessment capabilities before, during, and after an instructed task., Comment: Presented at the 2020 Workshop on Assessing, Explaining, and Conveying Robot Proficiency for Human-Robot Teaming

There is a close connection between health and the quality of one's social life. Strong social bonds are essential for health and wellbeing, but often health conditions can detrimentally affect a person's ability to interact with others. This can become a vicious cycle resulting in further decline in health. For this reason, the social management of health is an important aspect of healthcare. We propose that socially assistive robots (SARs) could help people with health conditions maintain positive social lives by supporting them in social interactions. This paper makes three contributions, as detailed below. We develop a framework of social mediation functions that robots could perform, motivated by the special social needs that people with health conditions have. In this framework we identify five types of functions that SARs could perform: a) changing how the person is perceived, b) enhancing the social behavior of the person, c) modifying the social behavior of others, d) providing structure for interactions, and e) changing how the person feels. We thematically organize and review the existing literature on robots supporting human-human interactions, in both clinical and non-clinical settings, and explain how the findings and design ideas from these studies can be applied to the functions identified in the framework. Finally, we point out and discuss challenges in designing SARs for supporting social interactions, and highlight opportunities for future robot design and HRI research on the mediator role of robots., Comment: 21 pages, 2 figs

We present an approach to generating natural language justifications of decisions derived from norm-based reasoning. Assuming an agent which maximally satisfies a set of rules specified in an object-oriented temporal logic, the user can ask factual questions (about the agent's rules, actions, and the extent to which the agent violated the rules) as well as "why" questions that require the agent comparing actual behavior to counterfactual trajectories with respect to these rules. To produce natural-sounding explanations, we focus on the subproblem of producing natural language clauses from statements in a fragment of temporal logic, and then describe how to embed these clauses into explanatory sentences. We use a human judgment evaluation on a testbed task to compare our approach to variants in terms of intelligibility, mental model and perceived trust., Comment: Accepted to the Proceedings of the 12th International Conference on Natural Language Generation (INLG 2019)

We present a set of capabilities allowing an agent planning with moral and social norms represented in temporal logic to respond to queries about its norms and behaviors in natural language, and for the human user to add and remove norms directly in natural language. The user may also pose hypothetical modifications to the agent's norms and inquire about their effects., Comment: Accepted to the 1st Workshop on Interactive Natural Language Technology for Explainable Artificial Intelligence (NL4XAI)

HRI researchers have made major strides in developing robotic architectures that are capable of reading a limited set of social cues and producing behaviors that enhance their likeability and feeling of comfort amongst humans. However, the cues in these models are fairly direct and the interactions largely dyadic. To capture the normative qualities of interaction more robustly, we propose consent as a distinct, critical area for HRI research. Convening important insights in existing HRI work around topics like touch, proxemics, gaze, and moral norms, the notion of consent reveals key expectations that can shape how a robot acts in social space. By sorting various kinds of consent through social and legal doctrine, we delineate empirical and technical questions to meet consent challenges faced in major application domains and robotic roles. Attention to consent could show, for example, how extraordinary, norm-violating actions can be justified by agents and accepted by those around them. We argue that operationalizing ideas from legal scholarship can better guide how robotic systems might cultivate and sustain proper forms of consent.

Human-robot communication in situated environments involves a complex interplay between knowledge representations across a wide variety of modalities. Crucially, linguistic information must be associated with representations of objects, locations, people, and goals, which may be represented in very different ways. In previous work, we developed a Consultant Framework that facilitates modality-agnostic access to information distributed across a set of heterogeneously represented knowledge sources. In this work, we draw inspiration from cognitive science to augment these distributed knowledge sources with Short Term Memory Buffers to create an STM-augmented algorithm for referring expression generation. We then discuss the potential performance benefits of this approach and insights from cognitive science that may inform future refinements in the design of our approach., Comment: International Conference on Social Robotics (ICSR) 2018

In this paper we describe moral quasi-dilemmas (MQDs): situations similar to moral dilemmas, but in which an agent is unsure whether exploring the plan space or the world may reveal a course of action that satisfies all moral requirements. We argue that artificial moral agents (AMAs) should be built to handle MQDs (in particular, by exploring the plan space rather than immediately accepting the inevitability of the moral dilemma), and that MQDs may be useful for evaluating AMA architectures., Comment: Accepted to the International Conference on Robot Ethics and Standards (ICRES), 2018

Recent work has addressed using formulas in linear temporal logic (LTL) as specifications for agents planning in Markov Decision Processes (MDPs). We consider the inverse problem: inferring an LTL specification from demonstrated behavior trajectories in MDPs. We formulate this as a multiobjective optimization problem, and describe state-based ("what actually happened") and action-based ("what the agent expected to happen") objective functions based on a notion of "violation cost". We demonstrate the efficacy of the approach by employing genetic programming to solve this problem in two simple domains., Comment: Accepted to the 56th IEEE Conference on Decision and Control (CDC 2017)

Among the many anticipated roles for robots in the future is that of being a human teammate. Aside from all the technological hurdles that have to be overcome with respect to hardware and control to make robots fit to work with humans, the added complication here is that humans have many conscious and subconscious expectations of their teammates - indeed, we argue that teaming is mostly a cognitive rather than physical coordination activity. This introduces new challenges for the AI and robotics community and requires fundamental changes to the traditional approach to the design of autonomy. With this in mind, we propose an update to the classical view of the intelligent agent architecture, highlighting the requirements for mental modeling of the human in the deliberative process of the autonomous agent. In this article, we outline briefly the recent efforts of ours, and others in the community, towards developing cognitive teammates along these guidelines.

Artificial agents will need to be aware of human moral and social norms, and able to use them in decision-making. In particular, artificial agents will need a principled approach to managing conflicting norms, which are common in human social interactions. Existing logic-based approaches suffer from normative explosion and are typically designed for deterministic environments; reward-based approaches lack principled ways of determining which normative alternatives exist in a given environment. We propose a hybrid approach, using Linear Temporal Logic (LTL) representations in Markov Decision Processes (MDPs), that manages norm conflicts in a systematic manner while accommodating domain stochasticity. We provide a proof-of-concept implementation in a simulated vacuum cleaning domain., Comment: New version of paper - new evaluations, accepted to AAAI 2018

Current measures of machine intelligence are either difficult to evaluate or lack the ability to test a robot's problem-solving capacity in open worlds. We propose a novel evaluation framework based on the formal notion of MacGyver Test which provides a practical way for assessing the resilience and resourcefulness of artificial agents.

Collaborative human activities are grounded in social and moral norms, which humans consciously and subconsciously use to guide and constrain their decision-making and behavior, thereby strengthening their interactions and preventing emotional and physical harm. This type of norm-based processing is also critical for robots in many human-robot interaction scenarios (e.g., when helping elderly and disabled persons in assisted living facilities, or assisting humans in assembly tasks in factories or even the space station). In this position paper, we will briefly describe how several components in an integrated cognitive architecture can be used to implement processes that are required for normative human-robot interactions, especially in collaborative tasks where actions and situations could potentially be perceived as threatening and thus need a change in course of action to mitigate the perceived threats., Comment: Presented at "2nd Workshop on Cognitive Architectures for Social Human-Robot Interaction 2016 (arXiv:1602.01868)"