How joshualikaist/motar works
MOTAR is a simulation research system for measuring how target motion, perception error, safety-filter geometry, and observation/rendering changes affect frozen UAV tracking policies. The documented evaluation loop runs from configured obstacle/target scenarios through camera and LiDAR sensing, perception and temporal state, policy action, speed limiting, vehicle control, and simulated outcomes. Research tooling and result records support the evidence workflow, but the sampled source does not establish every experiment-script connection; the graph therefore emphasizes documented runtime flow and omits uncertain wiring.
Main entry points: navrl_task.py Navigation Task · robot_manager.py Robot Manager · IGE_env_manager.py Environment Manager · sim_builder.py Simulator Builder
Components 20
Scenario Setup
- Task Config
aerial_gym/config/task_config/navrl_task_config.py - Environment Config
aerial_gym/config/env_config/navrl_bars_env.py - Robot Config
aerial_gym/config/robot_config/navrl_quad_config.py - Target Motion
aerial_gym/task/navrl_task/target_motion.py
Simulation Runtime
- Simulator Builder
aerial_gym/sim/sim_builder.py - Environment Manager
aerial_gym/env_manager/IGE_env_manager.py - Obstacle Placement
aerial_gym/env_manager/asset_manager.py - Robot Manager
aerial_gym/robots/robot_manager.py - Camera and LiDAR
aerial_gym/robots/robot_manager.py
Tracking Loop
- Navigation Task
aerial_gym/task/navrl_task/navrl_task.py - Perception
aerial_gym/task/navrl_task/navrl_perception.py - Track State
aerial_gym/task/navrl_task/navrl_search_state.py - Transformer Policy
aerial_gym/rl_training/rl_games/navrl_transformer_network.py - LiDAR Speed Filter
aerial_gym/task/navrl_task/speed_governor.py - Lee Controller
aerial_gym/control/controllers/base_lee_controller.py
Research Evidence
- Episode Outcome
aerial_gym/task/navrl_task/navrl_task.py - Perception Measurement
tools - Static Renderer
tools/renderer_validation/public_pipeline.py - Policy Training
aerial_gym/rl_training/rl_games/runner.py
Other components
- Researcher
Connections 22
- Researcher → Task Config: selects task
- Task Config → Simulator Builder: configures
- Environment Config → Simulator Builder: configures
- Robot Config → Robot Manager: configures
- Simulator Builder → Environment Manager: builds simulation
- Environment Manager → Obstacle Placement: places obstacles
- Environment Manager → Robot Manager: creates environments
- Robot Manager → Camera and LiDAR: provides sensor data
- Camera and LiDAR → Navigation Task: observations
- Navigation Task → Perception: processes observations
- Perception → Track State: updates tracks
- Navigation Task → Transformer Policy: requests action
- Navigation Task → LiDAR Speed Filter: limits command
- Navigation Task → Lee Controller: sends command
- Lee Controller → Robot Manager: controls vehicle
- Target Motion → Navigation Task: supplies target state
- Navigation Task → Episode Outcome: records episode
- Researcher → Perception Measurement: measures error
- Perception Measurement → Researcher: reports measurements
- Researcher → Static Renderer: runs render checks
- Researcher → Policy Training: trains policies
- Policy Training → Transformer Policy: produces policy