Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Animal-AI Environment: A Virtual Laboratory For Comparative Cognition and Artificial Intelligence Research

Konstantinos Voudouris, Ibrahim Alhas, Wout Schellaert, Matteo G. Mecattaf, Ben Slater, Matthew Crosby, Joel Holmes, John Burden, Niharika Chaubey, Niall Donnelly, Matishalin Patel, Marta Halina, José Hernández-Orallo, Lucy G. Cheke

TL;DR

The paper presents the Animal-AI Environment as a virtual laboratory designed to fuse artificial intelligence and comparative cognition research. It details a major update with features such as interactive dispensers, sign boards, richer graphics, and faster agent training, along with a library of thousands of tasks for comprehensive testing. It demonstrates the platform’s utility through experiments with Random, Heuristic, PPO, and Dreamer-v3 agents across foraging, operant chamber, and the Animal-AI Testbed, highlighting curricula and human-comparison findings. The work underscores the platform’s potential to drive interdisciplinary insights and outlines concrete directions for future enhancements and broader framework integration.

Abstract

The Animal-AI Environment is a unique game-based research platform designed to facilitate collaboration between the artificial intelligence and comparative cognition research communities. In this paper, we present the latest version of the Animal-AI Environment, outlining several major features that make the game more engaging for humans and more complex for AI systems. These features include interactive buttons, reward dispensers, and player notifications, as well as an overhaul of the environment's graphics and processing for significant improvements in agent training time and quality of the human player experience. We provide detailed guidance on how to build computational and behavioural experiments with the Animal-AI Environment. We present results from a series of agents, including the state-of-the-art deep reinforcement learning agent Dreamer-v3, on newly designed tests and the Animal-AI Testbed of 900 tasks inspired by research in the field of comparative cognition. The Animal-AI Environment offers a new approach for modelling cognition in humans and non-human animals, and for building biologically inspired artificial intelligence.

The Animal-AI Environment: A Virtual Laboratory For Comparative Cognition and Artificial Intelligence Research

TL;DR

The paper presents the Animal-AI Environment as a virtual laboratory designed to fuse artificial intelligence and comparative cognition research. It details a major update with features such as interactive dispensers, sign boards, richer graphics, and faster agent training, along with a library of thousands of tasks for comprehensive testing. It demonstrates the platform’s utility through experiments with Random, Heuristic, PPO, and Dreamer-v3 agents across foraging, operant chamber, and the Animal-AI Testbed, highlighting curricula and human-comparison findings. The work underscores the platform’s potential to drive interdisciplinary insights and outlines concrete directions for future enhancements and broader framework integration.

Abstract

The Animal-AI Environment is a unique game-based research platform designed to facilitate collaboration between the artificial intelligence and comparative cognition research communities. In this paper, we present the latest version of the Animal-AI Environment, outlining several major features that make the game more engaging for humans and more complex for AI systems. These features include interactive buttons, reward dispensers, and player notifications, as well as an overhaul of the environment's graphics and processing for significant improvements in agent training time and quality of the human player experience. We provide detailed guidance on how to build computational and behavioural experiments with the Animal-AI Environment. We present results from a series of agents, including the state-of-the-art deep reinforcement learning agent Dreamer-v3, on newly designed tests and the Animal-AI Testbed of 900 tasks inspired by research in the field of comparative cognition. The Animal-AI Environment offers a new approach for modelling cognition in humans and non-human animals, and for building biologically inspired artificial intelligence.
Paper Structure (30 sections, 16 figures, 6 tables)

This paper contains 30 sections, 16 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Collaborative Interdisciplinary Research with Animal-AI
  3. The Animal-AI Environment: A Comprehensive Overview
  4. Environment Design and Modes of Use
  5. Environment Specification
  6. The Arena & Agent
  7. Immovable Objects
  8. Movable Objects
  9. Valenced Objects
  10. Dispensers
  11. Sign Boards
  12. Physics Simulation
  13. Reward Signal & Health
  14. Using Animal-AI with Computational Models
  15. Using Animal-AI with Human Participants
  16. ...and 15 more sections

Figures (16)

  • Figure 1: The Animal-AI Environment is a research tool that facilitates a virtuous cycle between AI research and animal behaviour research.
  • Figure 2: An example of a task testing object permanence in the Animal-AI Environment. Left: Bird's-eye view of the arena, with the location of the agent indicated by the grey arrow. Right: The agent's view from its starting position. It is frozen while the rewarding green sphere drops down behind the blue wall (see voudouris2022evaluatingvoudouris2024investigating).
  • Figure 3: The Arena and the three Agent characters that can be selected for play and testing. Top: "hedgehog". Centre: "panda". Bottom: "pig".
  • Figure 4: The Immovable objects in the Animal-AI Environment. Top Left: CylinderTunnelTransparent. Top Right: CylinderTunnel. Bottom Left: Ramp. Bottom Centre: Wall. Bottom Right: WallTransparent.
  • Figure 5: The movable objects in the Animal-AI Environment. From Left to Right: LightBlock, HeavyBlock, UBlock, LBlock, JBlock, HollowBox.
  • ...and 11 more figures