Open-world AI requires artificial agents to be able to handle novelties that arise during task performance, i.e., agents must detect novelties and characterize them in order to be able to accommodate them effectively, especially in cases where sudden changes to the environment make task accomplishment impossible without utilizing the novelty. In this project, we are developing algorithms for all aspects of a formal framework and implementation thereof in a cognitive agent for novelty handling and demonstrate the efficacy of the proposed methods in various open-world tasks, including a crafting task in Minecraft.
We propose an integrated planning and learning approach that utilizes learning from failures and transferring knowledge over time to overcome novelty scenarios. The approach is more sample efficient in adapting to sudden and unknown changes (i.e., novelties) than the existing hybrid approaches. We showcase our results on a…
Keywords: reinforcement learning, planning, open-world AI, novelty accommodation
Speeding-up Continual Learning through Information Gains in Novel Experiences (2022)
We propose an integrated planning and learning approach that utilizes learning from failures and transferring knowledge over time to overcome novelty scenarios. The approach is more sample efficient in adapting to sudden and unknown changes (i.e., novelties) than the existing hybrid approaches. We showcase our results on a…
Keywords: reinforcement learning, planning, open-world AI, novelty accommodation
ACuTE: Automatic Curriculum Transfer from Simple to Complex Environments (2022)
We formulate the curriculum transfer problem, in which the schema of a curriculum optimized in a simpler, easy-to-solve environment (e.g., a grid world) is transferred to a complex, realistic scenario (e.g., a physics-based robotics simulation or the real world). We present "ACuTE", Automatic Curriculum Transfer from Simple to Complex…
Keywords: reinforcement learning, curriculum learning
BIPLEX: Creative Problem-Solving by Planning for Experimentation (2022)
We propose a novel problem solving algorithm called BIPLEX that involves hypothesis generation, experimentation, and outcome observation as part of the search for solutions. We introduce BIPLEX through various examples in a baking domain that demonstrate important features of the framework, including its representation of objects in…
Keywords: problem solving, open-world AI, planning, creativity
Analysis of the Future Potential of Autoencoders in Industrial Defect Detection (2022)
We investigated the anomaly detection behaviour of three convolutional autoencoder types - a "standard” convolutional autoencoder (CAE), a variational convolutional autoencoder (VAE) and an adversarial convolutional autoencoder (AAE) - by applying them to different visual anomaly detection scenarios. First, we utilized our three…
Keywords: vision, anomaly detection, defect detection
Autoencoders - A Comparative Analysis in the Realm of Anomaly Detection (2022)
We applied convolutional versions of a "standard" autoencoder (CAE), a variational autoencoder (VAE) and an adversarial autoencoder (AAE) to two different publicly available datasets and compared their anomaly detection performances. We used the MNIST dataset as a simple anomaly detection scenario. The CIFAR10 dataset was used to examine…
Keywords: vision, anomaly detection
The ability of AI agents and architectures to detect and adapt to sudden changes in their environments remains an outstand- ing challenge. In the context of multi-agent games, the agent may face novel situations where the rules of the game, the available actions, the environment dynamics, the behavior of other agents, as well as the…
As researchers are developing methods for detecting and accommo- dating novelties that will make AI agents more robust to unknown sudden changes in the “open worlds”, there is an increasing need for benchmark environments that allow for the systematic evalua- tions of the proposed AI techniques. We present “NovelGridworlds”, an OpenAI…
A Novelty-Centric Agent Architecture for Changing Worlds (2021)
Open-world AI requires artificial agents to cope with novelties that arise during task performance, i.e., they must (1) detect novelties, (2) characterize them, in order to (3) accommodate them, especially in cases where sudden changes to the environment make task accomplishment impossible without utilizing the novelty. We present a…
SPOTTER: Extending Symbolic Planning Operators through Targeted Reinforcement Learning (2021)
We developed a new integrated planning and reinforcement learning framework called SPOTTER to enable agents to solve problems unsolvable from their initial planning domain.