vocabulary: "1.0.0"

info:
  provider: "AI Habitat"
  description: "Taxonomy for Meta AI Habitat simulation framework covering simulation configuration, agent specifications, task definitions, and dataset resources for embodied AI research."
  created: "2026-04-19"
  modified: "2026-04-19"

operational:
  apis:
    - namespace: habitat-sim
      version: "3.0"
      baseUrl: https://github.com/facebookresearch/habitat-sim
      status: active
    - namespace: habitat-lab
      version: "3.0"
      baseUrl: https://github.com/facebookresearch/habitat-lab
      status: active

  resources:
    - name: SimulatorConfig
      apis: [habitat-sim]
      actions: [configure, initialize, reset, step]
      description: "Simulator configuration and lifecycle"
    - name: AgentConfig
      apis: [habitat-sim, habitat-lab]
      actions: [configure, spawn, observe, act]
      description: "Agent configuration including sensors and action space"
    - name: SensorSpec
      apis: [habitat-sim]
      actions: [configure, read]
      description: "Sensor specification for RGB, depth, semantic, and GPS sensors"
    - name: Episode
      apis: [habitat-lab]
      actions: [load, reset, evaluate]
      description: "Training and evaluation episode with start state and goal"
    - name: TaskConfig
      apis: [habitat-lab]
      actions: [configure, measure, evaluate]
      description: "Task type configuration and measurement specification"
    - name: NavigationGoal
      apis: [habitat-lab]
      actions: [set, reach]
      description: "Point-goal or object-goal specification for navigation tasks"
    - name: Observation
      apis: [habitat-sim, habitat-lab]
      actions: [get, process]
      description: "Multi-modal sensor observations from simulation step"

  actions:
    - name: step
      pattern: write
      description: "Advance simulation by one timestep with an action"
    - name: reset
      pattern: write
      description: "Reset environment to episode start state"
    - name: observe
      pattern: read
      description: "Get current sensor observations from agent"
    - name: configure
      pattern: write
      description: "Set configuration parameters"
    - name: evaluate
      pattern: read
      description: "Compute task metrics and success criteria"

  schemas:
    simulation:
      - name: SimulatorConfig
        description: "Top-level simulator configuration"
        keyFields: [scene_id, default_agent_id, enable_physics, gpu_device_id]
      - name: AgentConfig
        description: "Agent body and sensor configuration"
        keyFields: [height, radius, sensor_specifications, action_space]
      - name: SensorSpec
        description: "Individual sensor specification"
        keyFields: [uuid, sensor_type, resolution, hfov]
    tasks:
      - name: Episode
        description: "Single training/evaluation episode"
        keyFields: [episode_id, scene_id, start_position, goals]
      - name: NavigationGoal
        description: "Navigation goal with position and radius"
        keyFields: [position, radius]
      - name: TaskConfig
        description: "Task type and measurement configuration"
        keyFields: [type, max_episode_steps, measurements]
    observations:
      - name: Observation
        description: "Sensor observation dictionary"
        keyFields: [rgb, depth, semantic, gps, compass]
      - name: AgentObservation
        description: "Full agent observation dict with task-specific sensors"
        keyFields: [rgb, depth, pointgoal_with_gps_compass, objectgoal]

  enums:
    sensorTypes:
      - COLOR
      - DEPTH
      - SEMANTIC
      - EQUIRECTANGULAR
      - PANORAMIC
    taskTypes:
      - PointNav-v1
      - ObjectNav-v1
      - ImageNav-v1
      - RearrangePick-v0
      - RearrangePlacement-v0
      - HumanoidReach-v0
    robotTypes:
      - LocoBot
      - FetchRobot
      - FrankaArm
      - AlienGoQuadruped
    datasets:
      - HM3D
      - HM3D-Semantics
      - MatterPort3D
      - Gibson
      - Replica
      - ReplicaCAD
      - HSSD

  authentication:
    schemes:
      - type: none
        description: "Local Python library - no API authentication required"

capability:
  workflows:
    - name: "Embodied Navigation Training"
      description: "Train navigation agents using Habitat-Sim and Habitat-Lab across point-goal, object-goal, and image-goal tasks"
      apisCombined: [habitat-sim, habitat-lab]
      personas: [AI Researcher, Robotics Engineer]
      domains: [Simulation, Reinforcement Learning]

  personas:
    - id: ai-researcher
      name: "AI Researcher"
      description: "Researcher developing embodied AI algorithms using Habitat for training and benchmarking"
      workflows: ["Embodied Navigation Training"]
    - id: robotics-engineer
      name: "Robotics Engineer"
      description: "Engineer using Habitat for sim-to-real transfer research with physical robot systems"
      workflows: ["Embodied Navigation Training"]

  domains:
    - name: "Simulation"
      description: "High-performance 3D environment simulation and rendering"
      resources: [SimulatorConfig, AgentConfig, SensorSpec, Observation]
    - name: "Embodied AI Tasks"
      description: "Navigation, manipulation, and collaboration task definitions"
      resources: [Episode, TaskConfig, NavigationGoal]

crossReference:
  - resource: SimulatorConfig
    operations: [configure, initialize, reset]
    workflows: ["Embodied Navigation Training"]
    personas: [AI Researcher, Robotics Engineer]
  - resource: Episode
    operations: [load, reset, evaluate]
    workflows: ["Embodied Navigation Training"]
    personas: [AI Researcher]
  - resource: Observation
    operations: [get, process]
    workflows: ["Embodied Navigation Training"]
    personas: [AI Researcher, Robotics Engineer]