To Örebro University

In this short paper, we consider scenarios in human-robot collaboration where the robot relies on deliberation to activate communicative actions. We claim that reasoning about the perception of these actions is a key, but often disregarded ingredient for successful communication, and propose a pre-theoretical model that accounts for this. We also discuss the problems that may arise if perception is neglected when reasoning about communication.

In recent years, Human-AI Collaborative Systems (HAIC) has emerged as a key research frontier, with the aim of leveraging the complementary strengths of humans and artificial intelligence. Research in this area has demonstrated the potential of HAIC systems in a variety of fields, including healthcare, creative arts, finance, manufacturing, and education. These systems not only improve problem-solving and performance, but also enable novel forms of human–machine co-creation and decision-making. The HAIC workshop provides an interdisciplinary forum to explore these challenges and opportunities, fostering dialogue between communities towards a general framework for the design, evaluation and implementation of next-generation human-AI collaborative systems.

Automated planning (AI planning) solves complex sequential decision-making problems by searching for action sequences given a domain definition that models the actions a system can take. However, hand-crafting action models for complex problems is often challenging and cumbersome even for experts. To address this, we present Lifted Action Model Learning (LAML), a novel approach for learning planning domains from plan traces obtained by noisy observations of environment states. LAML integrates data lifting, decision-tree learning, and logic induction to derive abstract action models and generate planning domain representations. We evaluate our approach on 23 benchmark domains from the International Planning Competition, comparing its performance to state-of-the-art methods. Our evaluation considers multiple criteria, including domain reconstruction through comparison with reference domains, plan generation feasibility, comparison with historical plans, and plan validation success rate. Experimental results demonstrate that LAML not only reconstructs more accurate action models but also exhibits strong robustness to noise.

In this position paper, we argue for the need for deliberation in communication for artificial agents that perform tasks together with humans. Existing works use a set of terms and concepts with different meanings, resulting in ambiguity which does not allow for a general framework. As an initial step towards such a framework, we propose the notion of deliberative communication, clarify the necessary concepts and terminology, highlight the capabilities required in using deliberation for agents that communicate with human users, and discuss the main challenges.

In many robotic applications, especially those involving humans and the environment, linguistic and visual information must be processed jointly and bound together. Existing works either encode the image or the language into a subsymbolic space, like the CLIP model, or create a symbolic space of extracted information, like the object detection models. In this paper, we propose to describe images by nouns and modifiers and introduce a new embedded binding space where the linguistic and visual cues can effectively be bound. We investigate how state-of-the-art models perform in recognizing nouns and modifiers from images, and propose our method by introducing a dataset and CLIP-like recognition techniques based on transfer learning and metric learning. We show real-world experiments that demonstrate the practical applicability of our approach to robotics applications. Our results indicate that our method can surpass the state-of-the-art in recognizing nouns and modifiers from images. Interestingly, our method exhibits a language characteristic related to context sensitivity.

In recent years, Artificial Intelligence (AI) has gained increasing popularity in the area of art creation, by demonstrating its great potential. Research in this topic has developed AI systems able to generate creative outputs in fields such as music, painting, games, design and scientific discovery, either autonomously or in collaboration with humans. Therefore, AI also helped to analyze and study the mechanisms of creativity from a broader perspective: from the socio-anthropological to psychological, as well as cognitive impact of the autonomous creative processes of artificial intelligence. These advances are leading to new opportunities research perspectives, while also posing challenging questions related to authorship, integrity, bias and evaluation of AI artistic outputs. CREAI, the workshop on AI and creativity, tries to address these research lines and aims to provide a forum for the AI community to discuss problems, challenges and innovative approaches in the various sub-fields of AI and creativity.

One of the difficulties of using AI planners in industrial applications pertains to the complexity of writing planning domain models. These models are typically constructed by domain planning experts and can become increasingly difficult to codify for large applications. In this paper, we describe our ongoing research on a novel approach to automatically learn planning domains from previously executed traces using Behavior Trees as an intermediate human-readable structure. By involving human planning experts in the learning phase, our approach can benefit from their validation. This paper outlines the initial steps we have taken in this research, and presents the challenges we face in the future.

Autonomous agents embedded in a physical environment need the ability to correctly perceive the state of the environment from sensory data. In partially observable environments, certain properties can be perceived only in specific situations and from certain viewpoints that can be reached by the agent by planning and executing actions. For instance, to understand whether a cup is full of coffee, an agent, equipped with a camera, needs to turn on the light and look at the cup from the top. When the proper situations to perceive the desired properties are unknown, an agent needs to learn them and plan to get in such situations. In this paper, we devise a general method to solve this problem by evaluating the confidence of a neural network online and by using symbolic planning. We experimentally evaluate the proposed approach on several synthetic datasets, and show the feasibility of our approach in a real-world scenario that involves noisy perceptions and noisy actions on a real robot.

Behavior Trees (BTs) are a formalism increasingly used to control the execution of robotic systems. The strength of BTs resides in their compact, hierarchical and transparent representation. However, when used in practical applications transparency is often hindered by the introduction of implicit run-time relations between nodes, e.g., because of data dependencies or hardware-related ordering constraints. Manually verifying the correctness of a BT with respect to these hidden relations is a tedious and error-prone task. This paper presents a modular planning-based approach for automatically testing BTs offline at design time, to identify possible executions that may violate given data and ordering constraints and to exhibit traces of these executions to help debugging. Our approach supports both basic and advanced BT node types, e.g., supporting parallel behaviors, and can be extended with other node types as needed. We evaluate our approach on BTs used in a commercially deployed robotics system and on a large set of randomly generated trees showing that our approach scales to realistic sizes of more than 3000 nodes. 

Autonomous agents embedded in a physical environment need the ability to recognize objects and their properties from sensory data. Such a perceptual ability is often implemented by supervised machine learning models, which are pre-trained using a set of labelled data. In real-world, open-ended deployments, however, it is unrealistic to assume to have a pre-trained model for all possible environments. Therefore, agents need to dynamically learn/adapt/extend their perceptual abilities online, in an autonomous way, by exploring and interacting with the environment where they operate. This paper describes a way to do so, by exploiting symbolic planning. Specifically, we formalize the problem of automatically training a neural network to recognize object properties as a symbolic planning problem (using PDDL). We use planning techniques to produce a strategy for automating the training dataset creation and the learning process. Finally, we provide an experimental evaluation in both a simulated and a real environment, which shows that the proposed approach is able to successfully learn how to recognize new object properties.