Skip to content

Sim benchmark report

  • generated (UTC): 20261103T180315Z
  • git commit: 07f62c5 plus the runtime fixes committed as 8a32837
  • trials per world: 2
  • goal order: pickup,dropoff,home
  • wheel_mu: 1.0
  • nav2 params: mote_bringup/config/nav2_params.yaml
  • ROS_DOMAIN_ID: 28

office_world.sdf

  • map revision: 20260708T000623
  • map: mote_simulation/sim_home/sites/office_world/floors/ground/map/map.yaml
  • successful trials: 2/2
metric mean std min max CV
goal success rate 1.000 0.000 1.000 1.000 0.0%
time-to-goal mean (s) 85.3 0.3 85.0 85.6 0.4%
ATE rmse (m) 0.117 0.003 0.113 0.120 3.0%
ATE max (m) 0.335 0.040 0.295 0.374 11.8%
odom ATE rmse (m) 0.040 0.001 0.039 0.041 2.3%
odom ATE max (m) 0.095 0.001 0.094 0.096 0.9%
min clearance (m) 0.189 0.002 0.187 0.192 1.1%
mean clearance (m) 0.790 0.000 0.789 0.790 0.0%
linear jerk rms 3.15 0.01 3.13 3.16 0.4%
direction reversals 0.0 0.0 0.0 0.0 —
recoveries total 0.0 0.0 0.0 0.0 —
aborts 0.0 0.0 0.0 0.0 —
est path length (m) 48.45 0.35 48.10 48.80 0.7%
per-trial | trial | goals ok | ATE rmse (m) | odom ATE (m) | min clr (m) | recoveries | aborts | | --- | --- | --- | --- | --- | --- | --- | | 0 | 3/3 | 0.120 | 0.041 | 0.192 | 0 | 0 | | 1 | 3/3 | 0.113 | 0.039 | 0.187 | 0 | 0 |

Notes

  • Localization error is ATE (truth vs estimate) after a rigid SE(2) alignment — the SLAM map frame and the Gazebo world frame do not share a fixed transform, so alignment is required before differencing.
  • Ground truth is the robot's true pose bridged from Gazebo's PosePublisher (/model/mote/pose, gz.msgs.Pose → PoseStamped).
  • Recovery counts are distinct goal IDs seen on the behavior-server action status topics (/spin, /backup, /drive_on_heading, /wait) — a best-effort proxy for how often Nav2 recovered.
  • CV = coefficient of variation (std/mean); the run-to-run variance to weigh when comparing two configs.