Search

The Abstraction and Reasoning Corpus (ARC) is a general artificial intelligence benchmark that poses difficulties for pure machine learning methods due to its requirement for fluid intelligence with a focus on reasoning and abstraction. In this work, we introduce an ARC solver, Generalized Planning for Abstract Reasoning (GPAR). It casts an ARC problem as a generalized planning (GP) problem, where a solution is formalized as a planning program with pointers. We express each ARC problem using the standard Planning Domain Definition Language (PDDL) coupled with external functions representing object-centric abstractions. We show how to scale up GP solvers via domain knowledge specific to ARC in the form of restrictions over the actions model, predicates, arguments and valid structure of planning programs. Our experiments demonstrate that GPAR outperforms the state-of-the-art solvers on the object-centric tasks of the ARC, showing the effectiveness of GP and the expressiveness of PDDL to model ARC problems. The challenges provided by the ARC benchmark motivate research to advance existing GP solvers and understand new relations with other planning computational models. Code is available at github.com/you68681/GPAR.

The Model AI Assignments session seeks to gather and dis- seminate the best assignment designs of the Artificial In- telligence (AI) Education community. Recognizing that as- signments form the core of student learning experience, we here present abstracts of five AI assignments from the 2024 session that are easily adoptable, playfully engaging, and flexible for a variety of instructor needs. Assignment spec- ifications and supporting resources may be found at http://modelai.gettysburg.edu.

Goal Recognition (GR) is a research problem that studies ways to infer the goal of an intelligent agent based on its observed behavior and knowledge of the environment in which the agent operates. A common assumption of GR is that the environment is static. However, in many real-world scenarios, for example, recognizing customers’ preferences, it is necessary to recognize the goals of multiple agents or multiple goals of a single agent over an extended period. Therefore, it is reasonable to expect the environment to change throughout a series of goal recognition tasks. This paper presents three process mining-based solutions to the problem of adaptive GR in a changing environment implemented as different control strategies of a system for solving standard GR problems. As a standard GR system that gets controlled, we use the system grounded in process mining techniques, as it can adjust its internal GR mechanisms based on data collected while observing the operating agents. We evaluated our control strategies over synthetic and real-world datasets. The synthetic datasets were generated using the extended version of the Goal Recognition Amidst Changing Environments (GRACE) tool. The datasets account for different types of changes and drifts in the environment. The evaluation results demonstrate a trade-off between the GR performance over time and the effort invested in adaptations of the GR mechanisms of the system, showing that few well-planned adaptations can lead to a consistently high GR performance.

Planning under complex uncertainty often asks for plans that can adapt to changing future conditions. To inform plan development during this process, exploration methods have been used to explore the performance of candidate policies given uncertainties. Nevertheless, these methods hardly enable adaptation by themselves, so extra efforts are required to develop the final adaptive plans, hence compromising the overall decision-making efficiency. This paper introduces Reinforcement Learning (RL) that employs closed-loop control as a new exploration method that enables automated adaptive policy-making for planning under uncertainty. To investigate its performance, we compare RL with a widely-used exploration method, Multi-Objective Evolutionary Algorithm (MOEA), in two hypothetical problems via computational experiments. Our results indicate the complementarity of the two methods. RL makes better use of its exploration history, hence always providing higher efficiency and providing better policy robustness in the presence of parameter uncertainty. MOEA quantifies objective uncertainty in a more intuitive way, hence providing better robustness to objective uncertainty. These findings will help researchers choose appropriate methods in different applications.

The problem of goal recognition requests to automatically infer an accurate probability distribution over possible goals an autonomous agent is attempting to achieve in the environment. The state-of-the-art approaches for goal recognition operate under full knowledge of the environment and possible operations the agent can take. This knowledge, however, is often not available in real-world applications. Given historical observations of the agents' behaviors in the environment, we learn skill models that capture how the agents achieved the goals in the past. Next, given fresh observations of an agent, we infer their goals by diagnosing deviations between the observations and all the available skill models. We present a framework that serves as an outline for implementing such data-driven goal recognition systems and its instance system implemented using process mining techniques. The evaluations we conducted using our publicly available implementation confirm that the approach is well-defined, i.e., all system parameters impact its performance, has high accuracy over a wide range of synthetic and real-world domains, which is comparable with the more knowledge-demanding state-of-the-art approaches, and operates fast.

In multi-agent planning, preserving the agents' privacy has become an increasingly popular research topic. For preserving the agents' privacy, agents jointly compute a plan that achieves mutual goals by keeping certain information private to the individual agents. Unfortunately, this can severely restrict the accuracy of the heuristic functions used while searching for solutions. It has been recently shown that, for centralized planning, blind search algorithms such as width-based search can solve instances of many existing domains in low polynomial time when they feature atomic goals. Moreover, the performance of goal-oriented search can be improved by combining it with width-based search. In this paper, we investigate the usage of width-based search in the context of (decentralised) collaborative multi-agent privacy-preserving planning, addressing the challenges related to the agents' privacy and performance. In particular, we show that width-based search is a very effective approach over several benchmark domains, even when the search is driven by heuristics that roughly estimate the distance from goal states, computed without using the private information of other involved agents. Moreover, we show that the use of width-based techniques can significantly reduce the number of messages transmitted among the agents, better preserving their privacy and improving their performance. An experimental study presented in the paper analyses the effectiveness of our techniques, and compares them with the state-of-the-art of collaborative multi-agent planning.

A transhumeral prosthesis restores missing anatomical segments below the shoulder, including the hand. Active prostheses utilize real-valued, continuous sensor data to recognize patient target poses, or goals, and proactively move the artificial limb. Previous studies have examined how well the data collected in stationary poses, without considering the time steps, can help discriminate the goals. In this case study paper, we focus on using time series data from surface electromyography electrodes and kinematic sensors to sequentially recognize patients' goals. Our approach involves transforming the data into discrete events and training an existing process mining-based goal recognition system. Results from data collected in a virtual reality setting with ten subjects demonstrate the effectiveness of our proposed goal recognition approach, which achieves significantly better precision and recall than the state-of-the-art machine learning techniques and is less confident when wrong, which is beneficial when approximating smoother movements of prostheses.

Goal Recognition (GR) is a research problem that studies ways to infer the goal of an intelligent agent based on its observed behavior and knowledge of the environment. A common assumption of GR is that the underlying environment is stationary. However, in many real-world scenarios, it is necessary to recognize agents’ goals over extended periods. Therefore, it is reasonable to assume that the environment will change throughout a series of goal recognition tasks. This paper introduces the problem of continuous GR over a changing environment. The solution to this problem is a GR system capable of recognizing agents’ goals over an extended period where the environment in which the agents operate changes. To support the evaluation of candidate solutions to this new GR problem, in this paper, we present the Goal Recognition Amidst Changing Environments (GRACE) tool for generating instances of the new problem. Specifically, the tool can be configured to generate GR problems that account for different environmental changes and drifts. GRACE can generate a series of modified environments over discrete time steps and the data induced by agents operating in the environment while completing different goals.

Farm livestock identification and welfare assessment using non-invasive digital technology have gained interest in agriculture in the last decade, especially for accurate traceability. This study aimed to develop a face recognition system for dairy farm cows using advanced deep-learning models and computer vision techniques. This approach is non-invasive and potentially applicable to other farm animals of importance for identification and welfare assessment. The video analysis pipeline follows standard human face recognition systems made of four significant steps: (i) face detection, (ii) face cropping, (iii) face encoding, and (iv) face lookup. Three deep learning (DL) models were used within the analysis pipeline: (i) face detector, (ii) landmark predictor, and (iii) face encoder. All DL models were finetuned through transfer learning on a dairy cow dataset collected from a robotic dairy farm located in the Dookie campus at The University of Melbourne, Australia. Results showed that the accuracy across videos from 89 different dairy cows achieved an overall accuracy of 84%. The computer program developed may be deployed on edge devices, and it was tested on NVIDIA Jetson Nano board with a camera stream. Furthermore, it could be integrated into welfare assessment previously developed by our research group.

We investigate the application of state-of-the-art goal recognition techniques for behaviour recognition over complex continuous domains using model predictive control (MPC) for trajectory generation. We formally define the problem of kinodynamic behaviour recognition and establish a set of baseline behaviours and performance measures in the complex domain of unmanned aerial maneuvers. We evaluate how well our approach performs over a range of standard aerial maneuvers and representative initial configurations of varying complexity. The work also highlights future research directions in compound model-based behaviour recognition and team behaviour recognition where multiple agents may be acting simultaneously.

New and emerging technologies, especially those based on non-invasive video and thermal infrared cameras, can be readily tested on robotic milking facilities. In this research, implemented non-invasive computer vision methods to estimate cow's heart rate, respiration rate, and abrupt movements captured using RGB cameras and machine learning modelling to predict eye temperature, milk production and quality are presented. RGB and infrared thermal videos (IRTV) were acquired from cows using a robotic milking facility. Results from 102 different cows with replicates (n = 150) showed that an artificial neural network (ANN) model using only inputs from RGB cameras presented high accuracy (R = 0.96) in predicting eye temperature (°C), using IRTV as ground truth, daily milk productivity (kg-milk-day-1), cow milk productivity (kg-milk-cow-1), milk fat (%) and milk protein (%) with no signs of overfitting. The ANN model developed was deployed using an independent 132 cow samples obtained on different days, which also rendered high accuracy and was similar to the model development (R = 0.93). This model can be easily applied using affordable RGB camera systems to obtain all the proposed targets, including eye temperature, which can also be used to model animal welfare and biotic/abiotic stress. Furthermore, these models can be readily deployed in conventional dairy farms.

The problem of probabilistic goal recognition consists of automatically inferring a probability distribution over a range of possible goals of an autonomous agent based on the observations of its behavior. The state-of-the-art approaches for probabilistic goal recognition assume the full knowledge about the world the agent operates in and possible agent's operations in this world. In this paper, we propose a framework for solving the probabilistic goal recognition problem using process mining techniques for discovering models that describe the observed behavior and diagnosing deviations between the discovered models and observations. The framework imitates the principles of observational learning, one of the core mechanisms of social learning exhibited by humans, and relaxes the above assumptions. It has been implemented in a publicly available tool. The reported experimental results confirm the effectiveness and efficiency of the approach, both for rational and irrational agents' behaviors.

Bushfires are increasing in number and intensity due to climate change. A newly developed low-cost electronic nose (e-nose) was tested on wines made from grapevines exposed to smoke in field trials. E-nose readings were obtained from wines from five experimental treatments: (i) low-density smoke exposure (LS), (ii) high-density smoke exposure (HS), (iii) high-density smoke exposure with in-canopy misting (HSM), and two controls: (iv) control (C; no smoke treatment) and (v) control with in-canopy misting (CM; no smoke treatment). These e-nose readings were used as inputs for machine learning algorithms to obtain a classification model, with treatments as targets and seven neurons, with 97% accuracy in the classification of 300 samples into treatments as targets (Model 1). Models 2 to 4 used 10 neurons, with 20 glycoconjugates and 10 volatile phenols as targets, measured: in berries one hour after smoke (Model 2; R = 0.98; R2 = 0.95; b = 0.97); in berries at harvest (Model 3; R = 0.99; R2 = 0.97; b = 0.96); in wines (Model 4; R = 0.99; R2 = 0.98; b = 0.98). Model 5 was based on the intensity of 12 wine descriptors determined via a consumer sensory test (Model 5; R = 0.98; R2 = 0.96; b = 0.97). These models could be used by winemakers to assess near real-time smoke contamination levels and to implement amelioration strategies to minimize smoke taint in wines following bushfires

Sequential decision problems for real-world applications often need to be solved in real-time, requiring algorithms to perform well with a restricted computational budget. Widthbased lookaheads have shown state-of-the-art performance in classical planning problems as well as over the Atari games with tight budgets. In this work we investigate width-based lookaheads over Stochastic Shortest paths (SSP). We analyse why width-based algorithms perform poorly over SSP problems, and overcome these pitfalls proposing a method to estimate costs-to-go. We formalize width-based lookaheads as an instance of the rollout algorithm, give a definition of width for SSP problems and explain its sample complexity. Our experimental results over a variety of SSP benchmarks show the algorithm to outperform other state-of-the-art rollout algorithms such as UCT and RTDP

The automatic generation of realistic behaviour such as tactical intercepts for Unmanned Aerial Vehicles (UAV) in air combat is a challenging problem. State-of-the-art solutions propose handcrafted algorithms and heuristics whose performance depends heavily on the initial conditions and aerodynamic properties of the UAVs involved. This paper shows how to employ domain-independent planners, embedded into professional multi-agent simulations, to implement two-level Model Predictive Control (MPC) hybrid control systems for simulated UAVs. We compare the performance of controllers using planners with others based on behaviour trees that implement real world tactics. Our results indicate that hybrid planners derive novel and effective tactics from first principles inherent to the dynamical constraints UAVs are subject to.

We introduce a novel framework to formalize and solve transparent planning tasks by executing actions selected in a suitable and timely fashion. A transparent planning task is defined as a task where the objective of the agent is to communicate its true goal to observers, thereby making its intentions and its action selection transparent. We formally define and model these tasks as Goal Pomdps where the state space is the Cartesian product of the states of the world and a given set of hypothetical goals. Action effects are deterministic in the world states of the problem but probabilistic in the observer's beliefs. Transition probabilities are obtained from making a call to a model-based plan recognition algorithm, which we refer to as an observer stereotype. We propose an action selection strategy via online planning that seeks actions to quickly convey the goal being pursued to an observer assumed to fit a given stereotype. In order to keep run-times feasible, we propose a novel model-based plan recognition algorithm that approximates well-known probabilistic plan recognition methods. The resulting on-line planner, after being evaluated over a diverse set of domains and three different observer stereotypes, is found to convey goal information faster than purely goal-directed planners.

The automatic generation of realistic behaviour such as tactical intercepts for Unmanned Aerial Vehicles (UAV) in air combat is a challenging problem. State-of-the-art solutions propose hand–crafted algorithms and heuristics whose performance depends heavily on the initial conditions and specific aerodynamic characteristics of the UAVs involved. This demo shows the ability of domain–independent planners, embedded into simulators, to generate on–line, feed–forward, control signals that steer simulated aircraft as best suits the situation

Copyright 2016 AAAI, all rights reserved.). Operator-counting is a recently developed framework for analysing and integrating many state-ofthe- art heuristics for planning using Linear Programming. In cost-optimal planning only the objective value of these heuristics is traditionally used to guide the search. However the primal solution, i.e. the operator counts, contains useful information. We exploit this information using a SATbased approach which given an operator-count, either finds a valid plan; or generates a generalized landmark constraint violated by that count. We show that these generalized landmarks can be used to encode the perfect heuristic, h∗, as a Mixed Integer Program. Our most interesting experimental result is that finding or refuting a sequence for an operator-count is most often empirically efficient, enabling a novel and promising approach to planning based on Logic-Based Benders Decomposition (LBBD). This paper originally appeared at ICAPS 2015 and is reproduced with the permission of the Association for Artificial Intelligence ([Davies et al., 2015]