Description
According to the task board documentation, Zone 2 supports up to five SC ports distributed across two rails, which motivated us to build a more diverse SC dataset with >2 SC targets.
Reference:
- task_board_description.md states: “Supports up to five SC ports...” (Zone 2).
When using an SC data collection config that sets task_1.target_module_name: sc_port_2 and defines sc_rail_0..sc_rail_4, the task board still spawns only sc_port_0 and sc_port_1.
During task execution/scoring, aic_engine fails because TF frame task_board/sc_port_2/sc_port_base_link is missing.
Expected: sc_port_2 should be spawned (or config should be rejected early with a clear validation error).
Actual:
Runtime error during scoring: Transform between aic_world and task_board/sc_port_2/sc_port_base_link not found ... Invalid frame ID ... frame does not exist.
This indicates a mismatch between documented task-board capacity (up to 5 SC ports) and the currently exposed simulator/evaluation SC target modules.
Setup method
Docker + Distrobox + Pixi (recommended)
OS
Ubuntu 24.04.3 LTS
System specs
CPU: 12th Gen Intel(R) Core(TM) i9-12900KF (24 CPUs reported)
RAM: 31 GiB
GPU: NVIDIA GeForce RTX 3090 (24576 MiB)
GPU driver: 580.126.09
Steps to reproduce
- Generate expanded trial config from the SC five-port template.
trial_variations:
- name: trial_bug_sc_five_ports_target_2
overrides:
scene.task_board.pose.x: 0.17
scene.task_board.pose.y: 0.0
scene.task_board.pose.z: 1.14
scene.task_board.pose.roll: 0.0
scene.task_board.pose.pitch: 0.0
scene.task_board.pose.yaw: 3.0
scene.task_board.sc_rail_0.entity_present: true
scene.task_board.sc_rail_0.entity_name: sc_mount_0
scene.task_board.sc_rail_0.entity_pose.translation: 0.006
scene.task_board.sc_rail_0.entity_pose.roll: 0.0
scene.task_board.sc_rail_0.entity_pose.pitch: 0.0
scene.task_board.sc_rail_0.entity_pose.yaw: 0.0
scene.task_board.sc_rail_1.entity_present: true
scene.task_board.sc_rail_1.entity_name: sc_mount_1
scene.task_board.sc_rail_1.entity_pose.translation: 0.024
scene.task_board.sc_rail_1.entity_pose.roll: 0.0
scene.task_board.sc_rail_1.entity_pose.pitch: 0.0
scene.task_board.sc_rail_1.entity_pose.yaw: 0.0
scene.task_board.sc_rail_2.entity_present: true
scene.task_board.sc_rail_2.entity_name: sc_mount_2
scene.task_board.sc_rail_2.entity_pose.translation: 0.041
scene.task_board.sc_rail_2.entity_pose.roll: 0.0
scene.task_board.sc_rail_2.entity_pose.pitch: 0.0
scene.task_board.sc_rail_2.entity_pose.yaw: 0.0
- Start sim/model stack.
- In another terminal, run aic_engine.
Relevant log output
[ERROR] [aic_engine]: Transform between aic_world and task_board/sc_port_2/sc_port_base_link not found in the tf tree, error: Invalid frame ID "task_board/sc_port_2/sc_port_base_link" passed to canTransform argument source_frame - frame does not exist
Additional context
No response
Description
According to the task board documentation, Zone 2 supports up to five SC ports distributed across two rails, which motivated us to build a more diverse SC dataset with >2 SC targets.
Reference:
When using an SC data collection config that sets task_1.target_module_name: sc_port_2 and defines sc_rail_0..sc_rail_4, the task board still spawns only sc_port_0 and sc_port_1.
During task execution/scoring, aic_engine fails because TF frame task_board/sc_port_2/sc_port_base_link is missing.
Expected: sc_port_2 should be spawned (or config should be rejected early with a clear validation error).
Actual:
Runtime error during scoring: Transform between aic_world and task_board/sc_port_2/sc_port_base_link not found ... Invalid frame ID ... frame does not exist.
This indicates a mismatch between documented task-board capacity (up to 5 SC ports) and the currently exposed simulator/evaluation SC target modules.
Setup method
Docker + Distrobox + Pixi (recommended)
OS
Ubuntu 24.04.3 LTS
System specs
CPU: 12th Gen Intel(R) Core(TM) i9-12900KF (24 CPUs reported)
RAM: 31 GiB
GPU: NVIDIA GeForce RTX 3090 (24576 MiB)
GPU driver: 580.126.09
Steps to reproduce
Relevant log output
Additional context
No response