Hugging Face's logo

cccat6
/
FlowMo-WM

Robotics
PyTorch
world-model
visual-world-model
model-based-control
surface-vehicle
hidden-drift
Model card Files
xet
 Community
FlowMo-WM / experiments /reports /paper_report.md
cccat6's picture
cccat6
Rename traditional baseline artifacts as LOS controllers
cc396fd verified
preview code
|
raw
history blame contribute delete
81.3 kB

FlowMo Clean-Image Paper Experiment Report

Observation Setup

  • Input: clean top-down RGB image, no flow arrows, no goal marker, no velocity vector.
  • Image size: 160 x 160.
  • Boat visual scale: 2.5.
  • Training target: absolute normalized pose rollout plus shared current-pose observation head.
  • Formal experiments include only A-category learned WMs and B-category traditional non-WM controllers.

Parameter Counts

Method Total Parameters Main Components
FlowMo 663,964 encoder: 340,416, state_history: 75,648, context_history: 75,648, to_z: 30,960, to_c: 17,544, base_delta: 45,808, residual_delta: 46,448, decoder: 31,492
LeWorldModel 664,612 encoder: 471,584, to_z: 43,328, transition: 87,104, decoder: 62,596
PlaNet RSSM 664,644 encoder: 471,584, recurrent: 83,880, posterior: 39,108, prior: 27,588, decoder: 42,484
TD-MPC2 Dynamics 667,780 encoder: 471,584, history: 87,936, to_z: 30,960, transition: 45,808, decoder: 31,492

Prediction

Train, test, and planning use the same paper flow-family set. The test split uses independently sampled episodes and hidden flow conditions, and the report uses one unified flow-regime evaluation.

Paper Test Split

Method pos@10 pos@20 pos@40 pos@60 heading@20
FlowMo 0.107 0.151 0.245 0.344 0.072
FlowMo c=0 0.281 0.441 0.655 0.840 0.131
FlowMo shuffled c 0.353 0.563 0.861 1.100 0.166
LeWorldModel 0.173 0.246 0.364 0.479 0.106
PlaNet RSSM 0.161 0.223 0.331 0.426 0.075
TD-MPC2 Dynamics 0.155 0.224 0.334 0.435 0.089

FlowMo Latent Probes

Split Target z R2 c R2 z+c R2
test Object momentum 0.424 0.423 0.667
test Local flow -0.020 0.001 -0.015
test Episode drift -0.024 -0.000 -0.019

Probe interpretation: z is the short-history object-motion latent, c is the long-history ambient-drift context, and z+c is their concatenation. Probes are linear and are trained after freezing FlowMo.

Planning

reach_target_triangle_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.637 143.38 8.20 152.8
FlowMo shuffled 0.820 41/50 0.816 203.33 9.46 152.5
FlowMo zero 0.980 49/50 0.651 201.46 9.14 131.3
LeWorldModel inferred 0.960 48/50 0.812 266.07 10.41 175.1
PlaNet RSSM inferred 0.920 46/50 0.897 206.63 9.38 136.7
TD-MPC2 Dynamics inferred 0.940 47/50 0.733 275.98 11.04 152.4
PID/LOS controller inferred 1.000 50/50 0.628 78.93 7.96 222.1
No-Flow LOS Controller inferred 1.000 50/50 0.625 75.81 7.95 194.9
Current-Estimator LOS Controller inferred 1.000 50/50 0.626 79.29 7.92 195.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.622 75.64 7.92 194.7

reach_target_triangle_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.638 140.39 8.14 152.3
FlowMo shuffled 0.720 36/50 0.993 194.52 9.23 139.7
FlowMo zero 1.000 50/50 0.623 188.65 8.75 121.6
LeWorldModel inferred 0.940 47/50 0.815 243.11 9.21 153.5
PlaNet RSSM inferred 0.940 47/50 0.733 202.19 9.18 132.9
TD-MPC2 Dynamics inferred 0.980 49/50 0.663 224.67 9.45 128.6
PID/LOS controller inferred 1.000 50/50 0.628 75.04 7.93 214.9
No-Flow LOS Controller inferred 1.000 50/50 0.626 73.10 7.93 189.6
Current-Estimator LOS Controller inferred 1.000 50/50 0.631 75.88 7.90 189.5
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 72.96 7.91 189.3

reach_target_triangle_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.640 133.45 8.13 149.9
FlowMo shuffled 0.860 43/50 0.819 191.62 9.20 147.3
FlowMo zero 0.980 49/50 0.651 177.30 8.54 120.1
LeWorldModel inferred 0.900 45/50 0.815 273.30 10.38 170.0
PlaNet RSSM inferred 0.980 49/50 0.674 200.29 9.19 131.5
TD-MPC2 Dynamics inferred 1.000 50/50 0.620 221.98 9.41 126.9
PID/LOS controller inferred 1.000 50/50 0.627 77.83 7.88 212.0
No-Flow LOS Controller inferred 1.000 50/50 0.627 74.82 7.88 187.6
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 75.97 7.88 187.7
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 74.82 7.88 187.6

reach_target_triangle_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.639 141.81 8.15 146.6
FlowMo shuffled 0.860 43/50 0.896 201.23 9.49 146.3
FlowMo zero 0.980 49/50 0.656 204.15 9.42 134.2
LeWorldModel inferred 0.960 48/50 0.690 261.52 10.27 166.9
PlaNet RSSM inferred 0.960 48/50 0.677 206.13 9.48 134.1
TD-MPC2 Dynamics inferred 0.940 47/50 0.752 247.20 10.51 142.3
PID/LOS controller inferred 1.000 50/50 0.627 75.93 7.92 210.2
No-Flow LOS Controller inferred 1.000 50/50 0.620 72.79 7.91 185.8
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 75.60 7.90 186.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.621 73.82 7.90 186.1

reach_target_triangle_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.637 151.05 8.23 147.6
FlowMo shuffled 0.900 45/50 0.850 208.69 9.50 154.1
FlowMo zero 1.000 50/50 0.622 198.62 8.81 124.6
LeWorldModel inferred 0.980 49/50 0.658 265.91 10.06 173.2
PlaNet RSSM inferred 0.960 48/50 0.688 201.49 9.29 133.8
TD-MPC2 Dynamics inferred 0.980 49/50 0.686 233.72 9.87 133.6
PID/LOS controller inferred 1.000 50/50 0.629 72.03 7.92 212.3
No-Flow LOS Controller inferred 1.000 50/50 0.625 70.56 7.92 187.3
Current-Estimator LOS Controller inferred 1.000 50/50 0.631 73.78 7.90 187.5
Oracle-Flow LOS Controller inferred 1.000 50/50 0.622 70.58 7.91 187.3

reach_target_triangle_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.640 152.26 8.19 152.5
FlowMo shuffled 0.940 47/50 0.695 218.83 9.75 157.2
FlowMo zero 0.960 48/50 0.663 204.45 9.13 132.1
LeWorldModel inferred 0.880 44/50 1.110 243.57 9.67 157.4
PlaNet RSSM inferred 0.920 46/50 0.748 222.88 9.72 146.6
TD-MPC2 Dynamics inferred 0.960 48/50 0.702 219.43 9.21 128.0
PID/LOS controller inferred 1.000 50/50 0.627 76.58 7.91 222.4
No-Flow LOS Controller inferred 1.000 50/50 0.625 74.81 7.90 196.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 77.35 7.88 196.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.625 74.76 7.89 195.8

reach_target_triangle_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.639 146.04 8.19 154.2
FlowMo shuffled 0.840 42/50 0.826 200.45 9.13 144.2
FlowMo zero 1.000 50/50 0.624 200.55 9.06 129.4
LeWorldModel inferred 0.960 48/50 0.707 258.12 9.89 166.4
PlaNet RSSM inferred 0.940 47/50 0.711 215.27 9.61 143.3
TD-MPC2 Dynamics inferred 1.000 50/50 0.622 224.46 9.61 129.6
PID/LOS controller inferred 1.000 50/50 0.627 74.83 7.95 218.4
No-Flow LOS Controller inferred 1.000 50/50 0.628 73.15 7.94 192.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 75.67 7.91 191.8
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 72.90 7.91 191.5

reach_target_triangle_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.637 141.99 8.21 147.6
FlowMo shuffled 0.840 42/50 0.842 189.68 8.88 125.2
FlowMo zero 0.960 48/50 0.680 189.85 8.86 121.7
LeWorldModel inferred 0.920 46/50 0.993 247.63 9.83 159.6
PlaNet RSSM inferred 0.900 45/50 0.909 199.91 9.31 129.3
TD-MPC2 Dynamics inferred 0.980 49/50 0.685 237.48 10.05 131.7
PID/LOS controller inferred 1.000 50/50 0.626 74.17 7.91 210.6
No-Flow LOS Controller inferred 1.000 50/50 0.626 71.67 7.91 185.7
Current-Estimator LOS Controller inferred 1.000 50/50 0.626 73.99 7.89 185.9
Oracle-Flow LOS Controller inferred 1.000 50/50 0.623 71.56 7.90 185.6

reach_target_triangle_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.639 140.19 8.16 151.8
FlowMo shuffled 0.800 40/50 0.929 188.81 8.95 131.8
FlowMo zero 1.000 50/50 0.626 197.46 9.13 130.1
LeWorldModel inferred 0.960 48/50 0.805 257.52 9.99 161.9
PlaNet RSSM inferred 0.920 46/50 0.753 199.17 9.04 128.4
TD-MPC2 Dynamics inferred 0.940 47/50 0.681 268.33 10.91 146.8
PID/LOS controller inferred 1.000 50/50 0.629 78.46 8.00 218.5
No-Flow LOS Controller inferred 1.000 50/50 0.623 74.44 7.99 192.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 77.65 7.93 191.6
Oracle-Flow LOS Controller inferred 1.000 50/50 0.628 74.93 7.94 191.3

reach_target_triangle_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.637 153.08 8.19 152.3
FlowMo shuffled 0.940 47/50 0.677 214.61 9.59 154.0
FlowMo zero 0.980 49/50 0.656 205.23 8.85 129.2
LeWorldModel inferred 0.940 47/50 0.800 260.06 9.93 168.2
PlaNet RSSM inferred 0.940 47/50 0.709 223.40 9.79 146.6
TD-MPC2 Dynamics inferred 0.980 49/50 0.642 231.87 9.51 131.6
PID/LOS controller inferred 1.000 50/50 0.630 75.61 7.90 221.1
No-Flow LOS Controller inferred 1.000 50/50 0.625 73.99 7.90 195.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.629 76.37 7.89 195.2
Oracle-Flow LOS Controller inferred 1.000 50/50 0.623 74.03 7.89 195.1

reach_target_twin_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.668 152.86 9.07 175.3
FlowMo shuffled 0.760 38/50 1.319 191.98 9.63 149.8
FlowMo zero 0.680 34/50 1.525 213.27 10.17 147.9
LeWorldModel inferred 0.920 46/50 0.690 210.31 10.19 161.0
PlaNet RSSM inferred 0.860 43/50 1.027 228.46 11.40 175.5
TD-MPC2 Dynamics inferred 0.620 31/50 2.032 214.88 10.49 184.0
PID/LOS controller inferred 1.000 50/50 0.616 109.37 8.15 159.4
No-Flow LOS Controller inferred 0.960 48/50 0.680 120.71 8.39 154.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.623 116.27 8.31 154.6
Oracle-Flow LOS Controller inferred 0.980 49/50 0.656 120.74 8.40 154.4

reach_target_twin_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.800 142.29 8.88 165.7
FlowMo shuffled 0.660 33/50 1.388 209.65 10.10 171.4
FlowMo zero 0.660 33/50 1.535 217.92 10.34 154.0
LeWorldModel inferred 0.940 47/50 0.705 236.88 11.01 183.7
PlaNet RSSM inferred 0.920 46/50 1.039 258.79 12.81 200.5
TD-MPC2 Dynamics inferred 0.680 34/50 1.847 228.24 11.74 198.5
PID/LOS controller inferred 1.000 50/50 0.614 114.78 8.33 165.2
No-Flow LOS Controller inferred 0.780 39/50 0.899 126.99 8.81 165.2
Current-Estimator LOS Controller inferred 1.000 50/50 0.626 123.42 8.64 166.6
Oracle-Flow LOS Controller inferred 0.760 38/50 0.887 124.67 8.65 162.3

reach_target_twin_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.665 136.99 8.75 156.0
FlowMo shuffled 0.800 40/50 1.107 198.55 9.90 157.8
FlowMo zero 0.760 38/50 1.303 205.93 9.96 151.7
LeWorldModel inferred 0.960 48/50 0.689 217.51 10.66 167.8
PlaNet RSSM inferred 0.920 46/50 0.933 236.85 11.77 181.6
TD-MPC2 Dynamics inferred 0.720 36/50 1.552 244.78 12.21 205.3
PID/LOS controller inferred 1.000 50/50 0.611 105.99 8.12 156.5
No-Flow LOS Controller inferred 0.860 43/50 0.822 117.45 8.41 150.4
Current-Estimator LOS Controller inferred 1.000 50/50 0.617 114.19 8.33 153.1
Oracle-Flow LOS Controller inferred 0.860 43/50 0.822 117.45 8.41 150.4

reach_target_twin_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.636 135.40 8.69 155.3
FlowMo shuffled 0.700 35/50 1.293 216.25 10.52 170.3
FlowMo zero 0.680 34/50 1.421 207.95 10.24 152.8
LeWorldModel inferred 0.960 48/50 0.685 224.12 10.65 172.8
PlaNet RSSM inferred 0.900 45/50 0.817 238.39 12.01 188.0
TD-MPC2 Dynamics inferred 0.680 34/50 2.139 235.23 12.05 210.9
PID/LOS controller inferred 1.000 50/50 0.613 108.70 8.19 158.9
No-Flow LOS Controller inferred 0.780 39/50 0.875 117.54 8.39 149.5
Current-Estimator LOS Controller inferred 1.000 50/50 0.621 114.74 8.41 156.7
Oracle-Flow LOS Controller inferred 0.840 42/50 0.812 117.19 8.43 151.3

reach_target_twin_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.940 47/50 0.812 141.72 8.90 164.3
FlowMo shuffled 0.780 39/50 1.273 209.79 10.23 161.4
FlowMo zero 0.680 34/50 1.559 260.34 11.82 186.3
LeWorldModel inferred 0.800 40/50 0.937 224.23 10.57 171.4
PlaNet RSSM inferred 0.900 45/50 0.918 260.61 12.46 202.2
TD-MPC2 Dynamics inferred 0.660 33/50 2.090 225.78 11.67 196.7
PID/LOS controller inferred 1.000 50/50 0.611 115.17 8.37 167.1
No-Flow LOS Controller inferred 0.920 46/50 0.699 122.62 8.74 160.9
Current-Estimator LOS Controller inferred 1.000 50/50 0.623 119.95 8.66 166.6
Oracle-Flow LOS Controller inferred 0.940 47/50 0.684 122.64 8.74 161.1

reach_target_twin_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.666 150.14 8.87 172.8
FlowMo shuffled 0.740 37/50 1.182 231.62 10.80 191.0
FlowMo zero 0.660 33/50 1.694 237.63 10.85 172.8
LeWorldModel inferred 0.920 46/50 0.739 248.79 10.95 194.0
PlaNet RSSM inferred 0.940 47/50 0.835 243.89 11.78 189.2
TD-MPC2 Dynamics inferred 0.620 31/50 2.525 245.47 11.61 213.1
PID/LOS controller inferred 1.000 50/50 0.615 119.91 8.21 172.2
No-Flow LOS Controller inferred 0.820 41/50 0.804 127.22 8.55 166.4
Current-Estimator LOS Controller inferred 1.000 50/50 0.619 126.29 8.45 171.1
Oracle-Flow LOS Controller inferred 0.800 40/50 0.807 126.45 8.54 166.0

reach_target_twin_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.666 153.87 8.90 169.0
FlowMo shuffled 0.760 38/50 1.318 209.21 10.20 170.0
FlowMo zero 0.720 36/50 1.467 228.47 10.35 165.9
LeWorldModel inferred 0.940 47/50 0.715 233.55 10.63 178.2
PlaNet RSSM inferred 0.920 46/50 0.845 263.70 12.67 200.9
TD-MPC2 Dynamics inferred 0.720 36/50 2.291 239.02 12.08 209.5
PID/LOS controller inferred 0.960 48/50 0.638 119.09 8.32 172.2
No-Flow LOS Controller inferred 0.660 33/50 0.979 126.80 8.55 167.5
Current-Estimator LOS Controller inferred 1.000 50/50 0.626 124.69 8.58 170.9
Oracle-Flow LOS Controller inferred 0.680 34/50 0.966 127.77 8.56 171.1

reach_target_twin_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.693 131.86 8.68 154.2
FlowMo shuffled 0.820 41/50 1.110 198.38 9.89 160.7
FlowMo zero 0.680 34/50 1.510 225.34 10.72 161.4
LeWorldModel inferred 0.920 46/50 0.739 223.55 10.93 171.6
PlaNet RSSM inferred 0.920 46/50 0.942 235.32 11.92 183.6
TD-MPC2 Dynamics inferred 0.720 36/50 2.050 209.05 10.97 178.8
PID/LOS controller inferred 1.000 50/50 0.612 111.41 8.21 161.5
No-Flow LOS Controller inferred 0.860 43/50 0.797 119.41 8.46 154.6
Current-Estimator LOS Controller inferred 1.000 50/50 0.619 116.88 8.44 159.6
Oracle-Flow LOS Controller inferred 0.860 43/50 0.791 119.33 8.45 154.5

reach_target_twin_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.682 136.31 8.62 159.8
FlowMo shuffled 0.720 36/50 1.336 187.72 9.74 147.1
FlowMo zero 0.600 30/50 1.604 237.50 10.74 167.9
LeWorldModel inferred 0.940 47/50 0.671 214.64 10.42 165.0
PlaNet RSSM inferred 0.880 44/50 1.143 231.58 11.23 176.7
TD-MPC2 Dynamics inferred 0.700 35/50 1.817 227.47 11.63 195.0
PID/LOS controller inferred 0.980 49/50 0.634 111.31 8.29 160.9
No-Flow LOS Controller inferred 0.920 46/50 0.688 125.69 8.72 161.5
Current-Estimator LOS Controller inferred 1.000 50/50 0.625 118.62 8.45 157.5
Oracle-Flow LOS Controller inferred 0.860 43/50 0.697 115.29 8.32 147.8

reach_target_twin_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.665 142.88 8.76 163.4
FlowMo shuffled 0.720 36/50 1.125 202.16 9.80 164.2
FlowMo zero 0.600 30/50 1.846 226.53 10.37 164.9
LeWorldModel inferred 0.940 47/50 0.710 243.64 10.61 184.4
PlaNet RSSM inferred 0.880 44/50 0.988 237.61 11.66 188.2
TD-MPC2 Dynamics inferred 0.620 31/50 2.616 250.25 11.46 221.1
PID/LOS controller inferred 1.000 50/50 0.611 120.38 8.19 172.9
No-Flow LOS Controller inferred 0.800 40/50 0.849 128.83 8.50 168.8
Current-Estimator LOS Controller inferred 1.000 50/50 0.618 125.94 8.44 172.1
Oracle-Flow LOS Controller inferred 0.800 40/50 0.842 128.85 8.50 168.8

station_keeping_triangle_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.151 21.82 0.73 101.0
FlowMo shuffled 0.840 42/50 0.666 81.68 3.90 150.9
FlowMo zero 0.960 48/50 0.590 47.14 2.00 130.8
LeWorldModel inferred 1.000 50/50 0.388 50.84 1.59 101.4
PlaNet RSSM inferred 0.980 49/50 0.351 32.91 0.81 105.3
TD-MPC2 Dynamics inferred 1.000 50/50 0.292 33.72 1.20 101.5
PID/LOS controller inferred 1.000 50/50 0.104 88.74 0.77 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.114 73.74 0.91 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.130 76.19 0.96 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.244 51.66 1.12 101.0

station_keeping_triangle_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.155 22.95 0.72 101.0
FlowMo shuffled 0.840 42/50 0.727 90.24 4.11 154.0
FlowMo zero 1.000 50/50 0.577 45.17 1.94 126.6
LeWorldModel inferred 1.000 50/50 0.403 51.48 1.65 101.0
PlaNet RSSM inferred 1.000 50/50 0.379 29.21 0.72 103.2
TD-MPC2 Dynamics inferred 1.000 50/50 0.255 33.59 1.23 102.8
PID/LOS controller inferred 1.000 50/50 0.108 89.67 0.77 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.122 78.79 0.90 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.133 79.15 0.93 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.128 69.49 1.00 101.0

station_keeping_triangle_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.147 24.05 0.73 101.0
FlowMo shuffled 0.760 38/50 0.788 79.65 3.83 152.5
FlowMo zero 1.000 50/50 0.600 48.23 2.03 135.9
LeWorldModel inferred 1.000 50/50 0.393 50.38 1.61 101.4
PlaNet RSSM inferred 1.000 50/50 0.374 28.07 0.72 101.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.264 33.54 1.24 102.0
PID/LOS controller inferred 1.000 50/50 0.111 82.99 0.76 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.115 82.91 0.84 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.139 81.13 0.87 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.115 82.91 0.84 101.0

station_keeping_triangle_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.174 23.66 0.77 101.0
FlowMo shuffled 0.840 42/50 0.738 77.61 3.82 147.5
FlowMo zero 0.980 49/50 0.587 47.12 1.92 121.4
LeWorldModel inferred 1.000 50/50 0.377 49.02 1.61 101.1
PlaNet RSSM inferred 1.000 50/50 0.416 33.05 0.93 109.9
TD-MPC2 Dynamics inferred 1.000 50/50 0.305 36.20 1.35 104.0
PID/LOS controller inferred 1.000 50/50 0.105 97.92 0.78 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.121 78.26 0.93 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.148 78.39 0.98 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.259 76.55 1.35 101.0

station_keeping_triangle_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.148 24.57 0.75 101.0
FlowMo shuffled 0.860 43/50 0.769 81.48 3.87 147.6
FlowMo zero 0.980 49/50 0.591 45.42 1.95 127.0
LeWorldModel inferred 1.000 50/50 0.417 49.49 1.65 101.5
PlaNet RSSM inferred 1.000 50/50 0.386 27.92 0.73 101.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.299 37.25 1.28 104.1
PID/LOS controller inferred 1.000 50/50 0.096 84.87 0.77 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.107 81.05 0.88 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.128 81.33 0.93 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.125 70.22 0.96 101.0

station_keeping_triangle_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.143 22.49 0.71 101.0
FlowMo shuffled 0.760 38/50 0.749 74.79 3.31 147.1
FlowMo zero 0.980 49/50 0.570 46.28 1.85 131.1
LeWorldModel inferred 1.000 50/50 0.396 50.76 1.59 101.1
PlaNet RSSM inferred 1.000 50/50 0.378 29.34 0.67 101.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.294 34.15 1.17 103.0
PID/LOS controller inferred 1.000 50/50 0.100 93.71 0.74 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.121 82.22 0.87 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.125 81.96 0.92 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.142 71.75 1.02 101.0

station_keeping_triangle_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.149 22.88 0.71 101.0
FlowMo shuffled 0.820 41/50 0.681 65.53 3.26 142.5
FlowMo zero 1.000 50/50 0.579 45.99 1.97 134.1
LeWorldModel inferred 1.000 50/50 0.433 52.89 1.70 101.2
PlaNet RSSM inferred 1.000 50/50 0.374 30.71 0.72 103.8
TD-MPC2 Dynamics inferred 1.000 50/50 0.277 32.37 1.15 101.6
PID/LOS controller inferred 1.000 50/50 0.107 95.56 0.78 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.122 79.77 0.93 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.138 79.00 0.99 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.199 67.41 1.11 101.0

station_keeping_triangle_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.137 23.53 0.74 101.0
FlowMo shuffled 0.780 39/50 0.708 80.86 3.85 147.6
FlowMo zero 1.000 50/50 0.578 49.08 2.11 130.8
LeWorldModel inferred 1.000 50/50 0.390 50.59 1.70 101.3
PlaNet RSSM inferred 1.000 50/50 0.349 28.40 0.70 101.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.288 34.56 1.30 102.8
PID/LOS controller inferred 1.000 50/50 0.115 86.56 0.81 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.123 81.13 0.92 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.142 79.30 0.98 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.163 67.55 1.10 101.0

station_keeping_triangle_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.156 22.98 0.73 101.0
FlowMo shuffled 0.820 41/50 0.699 86.56 3.99 154.3
FlowMo zero 1.000 50/50 0.573 50.02 2.16 131.6
LeWorldModel inferred 1.000 50/50 0.398 50.04 1.61 101.6
PlaNet RSSM inferred 1.000 50/50 0.396 30.60 0.80 105.3
TD-MPC2 Dynamics inferred 1.000 50/50 0.266 33.60 1.24 101.8
PID/LOS controller inferred 1.000 50/50 0.105 98.15 0.77 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.140 75.08 0.92 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.140 75.92 0.96 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.218 63.30 1.10 101.0

station_keeping_triangle_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.147 22.56 0.71 101.0
FlowMo shuffled 0.720 36/50 0.848 71.14 3.31 141.6
FlowMo zero 1.000 50/50 0.596 43.72 1.83 126.3
LeWorldModel inferred 1.000 50/50 0.383 51.69 1.59 101.0
PlaNet RSSM inferred 1.000 50/50 0.378 29.75 0.69 102.4
TD-MPC2 Dynamics inferred 1.000 50/50 0.279 34.47 1.25 102.7
PID/LOS controller inferred 1.000 50/50 0.108 93.11 0.75 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.106 81.20 0.86 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.137 78.87 0.93 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.147 68.48 1.04 101.0

station_keeping_twin_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.258 31.98 0.48 101.0
FlowMo shuffled 0.900 45/50 0.623 65.41 1.52 123.9
FlowMo zero 1.000 50/50 0.371 80.36 1.42 118.6
LeWorldModel inferred 1.000 50/50 0.354 12.48 1.05 101.2
PlaNet RSSM inferred 1.000 50/50 0.565 77.82 1.43 136.3
TD-MPC2 Dynamics inferred 0.960 48/50 0.324 10.42 0.48 103.6
PID/LOS controller inferred 0.980 49/50 0.460 58.10 0.94 113.4
No-Flow LOS Controller inferred 0.780 39/50 0.533 32.02 0.71 107.0
Current-Estimator LOS Controller inferred 0.980 49/50 0.379 43.09 1.17 125.7
Oracle-Flow LOS Controller inferred 0.900 45/50 0.522 50.16 0.99 105.5

station_keeping_twin_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.206 29.57 0.47 101.6
FlowMo shuffled 0.920 46/50 0.556 52.63 1.47 114.0
FlowMo zero 1.000 50/50 0.355 69.35 1.13 114.7
LeWorldModel inferred 1.000 50/50 0.289 12.08 0.98 101.0
PlaNet RSSM inferred 0.980 49/50 0.464 67.81 1.13 122.9
TD-MPC2 Dynamics inferred 0.980 49/50 0.262 9.59 0.38 104.0
PID/LOS controller inferred 0.980 49/50 0.317 59.96 0.58 102.3
No-Flow LOS Controller inferred 1.000 50/50 0.330 31.29 0.56 104.2
Current-Estimator LOS Controller inferred 1.000 50/50 0.318 30.96 0.51 103.4
Oracle-Flow LOS Controller inferred 0.940 47/50 0.479 54.72 0.72 103.7

station_keeping_twin_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.190 32.77 0.41 101.0
FlowMo shuffled 0.880 44/50 0.623 64.38 1.88 118.9
FlowMo zero 1.000 50/50 0.381 80.52 1.18 115.2
LeWorldModel inferred 1.000 50/50 0.214 8.70 0.90 101.0
PlaNet RSSM inferred 1.000 50/50 0.315 52.67 0.57 101.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.088 5.58 0.20 101.0
PID/LOS controller inferred 1.000 50/50 0.261 38.31 0.32 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.285 10.61 0.29 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.283 10.89 0.28 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.285 10.61 0.29 101.0

station_keeping_twin_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.326 36.01 0.65 101.9
FlowMo shuffled 0.940 47/50 0.638 75.60 2.02 127.3
FlowMo zero 0.960 48/50 0.468 89.30 2.09 125.2
LeWorldModel inferred 0.980 49/50 0.487 26.72 1.38 116.6
PlaNet RSSM inferred 0.960 48/50 0.498 65.54 1.21 115.2
TD-MPC2 Dynamics inferred 0.980 49/50 0.480 20.39 0.91 122.6
PID/LOS controller inferred 0.980 49/50 0.422 80.02 1.05 106.6
No-Flow LOS Controller inferred 0.820 41/50 0.581 43.25 0.88 105.9
Current-Estimator LOS Controller inferred 0.900 45/50 0.479 47.28 0.96 108.2
Oracle-Flow LOS Controller inferred 0.700 35/50 0.794 56.40 1.04 105.4

station_keeping_twin_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.180 31.00 0.43 101.0
FlowMo shuffled 0.900 45/50 0.557 62.30 1.77 120.4
FlowMo zero 1.000 50/50 0.382 74.10 1.19 116.2
LeWorldModel inferred 1.000 50/50 0.223 9.09 0.92 101.0
PlaNet RSSM inferred 1.000 50/50 0.342 54.38 0.60 104.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.137 6.40 0.24 101.0
PID/LOS controller inferred 1.000 50/50 0.289 48.12 0.44 101.0
No-Flow LOS Controller inferred 1.000 50/50 0.284 20.26 0.34 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.284 20.72 0.34 101.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.355 42.30 0.50 101.2

station_keeping_twin_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.222 29.77 0.45 101.0
FlowMo shuffled 0.960 48/50 0.568 82.48 2.00 124.6
FlowMo zero 1.000 50/50 0.337 62.91 1.08 113.1
LeWorldModel inferred 1.000 50/50 0.326 12.33 0.97 105.4
PlaNet RSSM inferred 0.960 48/50 0.512 65.63 1.01 118.7
TD-MPC2 Dynamics inferred 0.980 49/50 0.331 10.50 0.43 103.8
PID/LOS controller inferred 1.000 50/50 0.362 71.34 0.69 103.3
No-Flow LOS Controller inferred 0.960 48/50 0.397 37.15 0.59 102.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.369 40.88 0.67 107.8
Oracle-Flow LOS Controller inferred 0.940 47/50 0.485 59.96 0.73 102.4

station_keeping_twin_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.271 33.14 0.56 101.0
FlowMo shuffled 0.900 45/50 0.605 72.32 2.04 128.4
FlowMo zero 1.000 50/50 0.357 64.56 1.16 112.7
LeWorldModel inferred 1.000 50/50 0.362 15.94 1.09 105.1
PlaNet RSSM inferred 0.960 48/50 0.488 70.55 1.15 127.9
TD-MPC2 Dynamics inferred 0.960 48/50 0.428 17.13 0.65 119.5
PID/LOS controller inferred 0.980 49/50 0.392 77.90 0.86 106.0
No-Flow LOS Controller inferred 0.880 44/50 0.515 43.16 0.74 104.6
Current-Estimator LOS Controller inferred 0.980 49/50 0.407 48.91 0.95 114.0
Oracle-Flow LOS Controller inferred 0.760 38/50 0.732 58.75 0.85 104.0

station_keeping_twin_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.243 32.29 0.49 101.0
FlowMo shuffled 0.960 48/50 0.469 55.00 1.52 119.5
FlowMo zero 1.000 50/50 0.404 74.09 1.25 117.3
LeWorldModel inferred 1.000 50/50 0.346 15.79 1.06 103.9
PlaNet RSSM inferred 0.960 48/50 0.454 68.90 1.15 129.9
TD-MPC2 Dynamics inferred 0.980 49/50 0.299 10.24 0.44 106.8
PID/LOS controller inferred 1.000 50/50 0.333 63.37 0.62 101.0
No-Flow LOS Controller inferred 0.980 49/50 0.364 31.45 0.55 101.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.340 32.65 0.57 103.0
Oracle-Flow LOS Controller inferred 0.900 45/50 0.560 52.19 0.77 106.4

station_keeping_twin_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.297 32.81 0.57 101.4
FlowMo shuffled 0.900 45/50 0.595 55.18 1.37 116.4
FlowMo zero 1.000 50/50 0.411 69.29 1.20 113.0
LeWorldModel inferred 1.000 50/50 0.383 18.26 1.20 109.0
PlaNet RSSM inferred 0.980 49/50 0.517 71.50 1.35 128.7
TD-MPC2 Dynamics inferred 0.900 45/50 0.468 16.87 0.68 122.5
PID/LOS controller inferred 1.000 50/50 0.394 78.08 0.86 101.4
No-Flow LOS Controller inferred 0.880 44/50 0.528 45.36 0.80 102.6
Current-Estimator LOS Controller inferred 1.000 50/50 0.426 49.39 0.95 110.7
Oracle-Flow LOS Controller inferred 0.820 41/50 0.642 59.56 0.97 103.2

station_keeping_twin_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.224 30.70 0.47 101.0
FlowMo shuffled 0.920 46/50 0.587 66.52 1.52 116.5
FlowMo zero 1.000 50/50 0.318 63.22 1.01 111.0
LeWorldModel inferred 1.000 50/50 0.311 13.17 0.96 102.4
PlaNet RSSM inferred 0.900 45/50 0.506 64.26 0.99 122.0
TD-MPC2 Dynamics inferred 1.000 50/50 0.292 10.83 0.42 104.1
PID/LOS controller inferred 0.980 49/50 0.312 65.40 0.57 101.0
No-Flow LOS Controller inferred 0.960 48/50 0.345 33.74 0.49 101.0
Current-Estimator LOS Controller inferred 0.980 49/50 0.334 35.41 0.51 102.8
Oracle-Flow LOS Controller inferred 0.920 46/50 0.490 54.58 0.66 101.5

waypoint_square_triangle_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.635 300.34 15.07 271.1
FlowMo shuffled 0.880 44/50 0.825 559.30 22.28 364.3
FlowMo zero 0.960 48/50 0.770 497.34 19.76 303.0
LeWorldModel inferred 0.980 49/50 0.666 455.91 19.28 302.2
PlaNet RSSM inferred 0.980 49/50 0.668 327.23 16.84 270.2
TD-MPC2 Dynamics inferred 0.960 48/50 0.735 327.10 16.15 244.9
PID/LOS controller inferred 1.000 50/50 0.629 190.59 15.36 420.1
No-Flow LOS Controller inferred 1.000 50/50 0.628 168.15 15.52 373.2
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 188.54 14.84 375.6
Oracle-Flow LOS Controller inferred 1.000 50/50 0.621 168.09 15.41 372.8

waypoint_square_triangle_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.634 312.65 15.41 277.5
FlowMo shuffled 0.900 45/50 0.797 566.29 22.69 377.2
FlowMo zero 0.880 44/50 0.997 526.32 20.98 320.7
LeWorldModel inferred 0.940 47/50 0.758 432.27 18.24 276.2
PlaNet RSSM inferred 1.000 50/50 0.627 332.90 16.99 282.6
TD-MPC2 Dynamics inferred 0.980 49/50 0.692 342.40 16.59 254.5
PID/LOS controller inferred 1.000 50/50 0.628 185.09 15.24 421.1
No-Flow LOS Controller inferred 1.000 50/50 0.630 166.23 15.46 374.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 181.89 14.87 374.4
Oracle-Flow LOS Controller inferred 1.000 50/50 0.625 165.73 15.41 372.8

waypoint_square_triangle_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.635 304.91 15.34 272.0
FlowMo shuffled 0.840 42/50 0.945 504.59 21.29 331.4
FlowMo zero 0.860 43/50 0.989 496.03 20.04 289.6
LeWorldModel inferred 0.940 47/50 0.761 462.02 19.37 300.8
PlaNet RSSM inferred 1.000 50/50 0.626 334.86 17.16 275.0
TD-MPC2 Dynamics inferred 0.980 49/50 0.699 335.82 16.62 252.3
PID/LOS controller inferred 1.000 50/50 0.629 183.71 15.19 411.4
No-Flow LOS Controller inferred 1.000 50/50 0.626 164.91 15.44 367.4
Current-Estimator LOS Controller inferred 1.000 50/50 0.630 179.09 14.86 368.1
Oracle-Flow LOS Controller inferred 1.000 50/50 0.626 164.91 15.44 367.4

waypoint_square_triangle_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.632 292.75 15.20 260.1
FlowMo shuffled 0.840 42/50 0.958 524.63 21.29 345.5
FlowMo zero 0.820 41/50 1.173 472.40 19.50 272.7
LeWorldModel inferred 0.960 48/50 0.721 422.39 18.33 270.2
PlaNet RSSM inferred 0.980 49/50 0.692 333.59 17.08 275.9
TD-MPC2 Dynamics inferred 1.000 50/50 0.621 336.89 16.76 245.5
PID/LOS controller inferred 1.000 50/50 0.630 189.33 15.31 407.8
No-Flow LOS Controller inferred 1.000 50/50 0.625 166.03 15.52 363.9
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 182.90 14.90 365.3
Oracle-Flow LOS Controller inferred 1.000 50/50 0.626 169.43 15.48 365.2

waypoint_square_triangle_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.635 307.72 15.10 277.5
FlowMo shuffled 0.780 39/50 1.124 573.11 23.31 391.3
FlowMo zero 0.940 47/50 0.790 508.29 20.18 303.4
LeWorldModel inferred 0.920 46/50 0.821 447.05 18.90 284.1
PlaNet RSSM inferred 1.000 50/50 0.626 327.61 16.69 269.3
TD-MPC2 Dynamics inferred 1.000 50/50 0.628 329.99 16.18 246.1
PID/LOS controller inferred 1.000 50/50 0.627 186.08 15.16 428.1
No-Flow LOS Controller inferred 1.000 50/50 0.626 166.74 15.39 378.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 181.14 14.84 378.9
Oracle-Flow LOS Controller inferred 1.000 50/50 0.628 165.98 15.36 377.2

waypoint_square_triangle_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.687 297.82 14.93 266.8
FlowMo shuffled 0.840 42/50 0.934 514.30 20.62 324.8
FlowMo zero 0.920 46/50 0.852 516.57 20.00 308.5
LeWorldModel inferred 0.960 48/50 0.712 451.51 18.92 293.4
PlaNet RSSM inferred 1.000 50/50 0.626 346.42 17.15 292.4
TD-MPC2 Dynamics inferred 1.000 50/50 0.624 347.88 16.52 262.3
PID/LOS controller inferred 1.000 50/50 0.630 189.25 15.21 432.5
No-Flow LOS Controller inferred 1.000 50/50 0.630 169.75 15.42 384.2
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 182.95 14.84 382.7
Oracle-Flow LOS Controller inferred 1.000 50/50 0.629 169.27 15.38 383.5

waypoint_square_triangle_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.633 298.89 14.90 265.9
FlowMo shuffled 0.900 45/50 0.734 555.15 22.62 367.9
FlowMo zero 0.900 45/50 0.957 519.70 20.66 307.2
LeWorldModel inferred 0.920 46/50 0.797 420.63 18.22 270.6
PlaNet RSSM inferred 1.000 50/50 0.630 326.79 16.80 270.5
TD-MPC2 Dynamics inferred 0.980 49/50 0.671 324.62 16.07 240.1
PID/LOS controller inferred 1.000 50/50 0.630 183.92 15.16 419.2
No-Flow LOS Controller inferred 1.000 50/50 0.627 164.87 15.36 372.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.632 180.92 14.79 373.1
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 165.49 15.35 372.0

waypoint_square_triangle_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.634 295.94 15.13 266.5
FlowMo shuffled 0.780 39/50 1.088 531.52 22.10 356.5
FlowMo zero 0.940 47/50 0.808 504.92 20.62 303.2
LeWorldModel inferred 0.940 47/50 0.776 429.31 18.42 271.6
PlaNet RSSM inferred 1.000 50/50 0.624 314.56 16.76 265.7
TD-MPC2 Dynamics inferred 0.980 49/50 0.680 318.19 16.17 238.6
PID/LOS controller inferred 1.000 50/50 0.630 184.94 15.18 408.3
No-Flow LOS Controller inferred 1.000 50/50 0.625 164.02 15.40 363.5
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 179.41 14.81 364.5
Oracle-Flow LOS Controller inferred 1.000 50/50 0.624 163.14 15.38 362.8

waypoint_square_triangle_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.636 304.20 15.16 278.9
FlowMo shuffled 0.880 44/50 0.864 540.25 22.03 349.9
FlowMo zero 0.940 47/50 0.810 508.09 20.12 313.1
LeWorldModel inferred 0.920 46/50 0.796 451.77 18.64 289.3
PlaNet RSSM inferred 0.960 48/50 0.710 331.05 16.85 273.1
TD-MPC2 Dynamics inferred 1.000 50/50 0.623 328.75 16.20 248.1
PID/LOS controller inferred 1.000 50/50 0.628 191.95 15.39 430.3
No-Flow LOS Controller inferred 1.000 50/50 0.628 170.69 15.57 380.6
Current-Estimator LOS Controller inferred 1.000 50/50 0.629 186.46 14.87 378.8
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 169.48 15.45 377.9

waypoint_square_triangle_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.680 313.26 15.24 270.0
FlowMo shuffled 0.780 39/50 1.230 540.76 21.36 344.9
FlowMo zero 0.960 48/50 0.721 519.64 20.32 311.2
LeWorldModel inferred 0.860 43/50 1.011 454.36 18.62 290.8
PlaNet RSSM inferred 1.000 50/50 0.623 349.14 17.30 288.1
TD-MPC2 Dynamics inferred 0.960 48/50 0.734 346.02 16.34 257.2
PID/LOS controller inferred 1.000 50/50 0.630 191.28 15.16 426.4
No-Flow LOS Controller inferred 1.000 50/50 0.622 169.69 15.39 379.8
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 185.48 14.84 379.8
Oracle-Flow LOS Controller inferred 1.000 50/50 0.629 170.17 15.36 379.5

waypoint_square_twin_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.780 325.51 15.55 343.7
FlowMo shuffled 0.640 32/50 1.678 602.47 25.48 488.4
FlowMo zero 0.540 27/50 1.740 695.99 29.02 510.3
LeWorldModel inferred 0.860 43/50 1.054 586.78 26.23 488.0
PlaNet RSSM inferred 0.600 30/50 1.961 511.69 21.93 446.6
TD-MPC2 Dynamics inferred 0.940 47/50 0.741 471.78 22.03 441.7
PID/LOS controller inferred 0.640 32/50 1.009 281.67 19.15 368.2
No-Flow LOS Controller inferred 0.060 3/50 1.691 298.34 18.67 370.3
Current-Estimator LOS Controller inferred 0.820 41/50 0.684 286.97 18.34 414.6
Oracle-Flow LOS Controller inferred 0.080 4/50 1.655 283.33 18.67 352.0

waypoint_square_twin_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.940 47/50 0.824 340.28 16.07 358.6
FlowMo shuffled 0.600 30/50 1.797 638.66 26.14 509.1
FlowMo zero 0.520 26/50 1.868 740.33 30.42 551.0
LeWorldModel inferred 0.820 41/50 1.374 551.49 24.66 463.6
PlaNet RSSM inferred 0.640 32/50 1.686 551.37 22.70 489.3
TD-MPC2 Dynamics inferred 0.900 45/50 0.906 473.83 21.97 461.2
PID/LOS controller inferred 0.440 22/50 0.976 311.95 19.44 400.8
No-Flow LOS Controller inferred 0.040 2/50 1.709 272.51 22.07 343.0
Current-Estimator LOS Controller inferred 0.900 45/50 0.662 312.56 19.24 462.4
Oracle-Flow LOS Controller inferred 0.020 1/50 1.742 301.03 23.37 380.0

waypoint_square_twin_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.775 315.63 15.58 336.6
FlowMo shuffled 0.640 32/50 1.744 567.72 23.29 453.0
FlowMo zero 0.620 31/50 1.622 697.94 29.49 527.2
LeWorldModel inferred 0.820 41/50 1.286 571.96 25.59 473.9
PlaNet RSSM inferred 0.680 34/50 1.723 495.36 21.25 422.6
TD-MPC2 Dynamics inferred 0.920 46/50 0.822 508.64 23.21 473.5
PID/LOS controller inferred 0.500 25/50 0.949 271.67 18.52 358.8
No-Flow LOS Controller inferred 0.020 1/50 1.725 251.51 18.44 414.0
Current-Estimator LOS Controller inferred 0.800 40/50 0.683 279.11 18.13 411.2
Oracle-Flow LOS Controller inferred 0.020 1/50 1.725 251.51 18.44 414.0

waypoint_square_twin_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.637 329.80 15.74 341.6
FlowMo shuffled 0.460 23/50 2.363 647.74 27.14 516.3
FlowMo zero 0.600 30/50 1.753 691.40 29.12 511.0
LeWorldModel inferred 0.880 44/50 1.092 556.59 25.06 465.5
PlaNet RSSM inferred 0.640 32/50 1.870 500.10 21.99 432.7
TD-MPC2 Dynamics inferred 0.980 49/50 0.669 475.40 22.15 437.8
PID/LOS controller inferred 0.720 36/50 1.101 287.63 19.59 371.2
No-Flow LOS Controller inferred 0.040 2/50 2.095 285.87 18.85 360.0
Current-Estimator LOS Controller inferred 0.840 42/50 0.999 292.41 19.30 428.0
Oracle-Flow LOS Controller inferred 0.020 1/50 1.963 311.82 19.21 394.0

waypoint_square_twin_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.638 303.67 15.25 324.2
FlowMo shuffled 0.540 27/50 1.950 572.38 24.38 461.9
FlowMo zero 0.540 27/50 1.864 602.16 25.94 445.7
LeWorldModel inferred 0.920 46/50 0.851 541.42 24.71 452.6
PlaNet RSSM inferred 0.780 39/50 1.330 522.76 22.29 443.5
TD-MPC2 Dynamics inferred 0.940 47/50 0.744 482.12 22.17 445.9
PID/LOS controller inferred 0.680 34/50 1.071 296.34 19.80 380.0
No-Flow LOS Controller inferred 0.080 4/50 2.008 288.44 21.05 363.0
Current-Estimator LOS Controller inferred 0.960 48/50 0.801 306.38 19.08 433.3
Oracle-Flow LOS Controller inferred 0.080 4/50 1.942 267.57 20.17 339.8

waypoint_square_twin_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.940 47/50 0.828 334.90 15.78 348.9
FlowMo shuffled 0.740 37/50 1.296 629.49 25.94 492.1
FlowMo zero 0.660 33/50 1.479 754.73 30.22 554.8
LeWorldModel inferred 0.880 44/50 1.133 584.54 24.72 493.9
PlaNet RSSM inferred 0.580 29/50 1.806 519.53 21.49 447.2
TD-MPC2 Dynamics inferred 0.940 47/50 0.757 536.15 22.87 491.4
PID/LOS controller inferred 0.540 27/50 0.940 320.62 19.45 406.7
No-Flow LOS Controller inferred 0.060 3/50 1.705 378.46 21.75 510.3
Current-Estimator LOS Controller inferred 0.740 37/50 0.766 291.79 18.19 427.9
Oracle-Flow LOS Controller inferred 0.040 2/50 1.860 373.12 19.21 492.0

waypoint_square_twin_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.691 368.48 17.10 376.2
FlowMo shuffled 0.540 27/50 1.987 597.76 24.11 484.9
FlowMo zero 0.580 29/50 1.736 717.53 29.41 543.3
LeWorldModel inferred 0.740 37/50 1.571 529.91 24.36 449.5
PlaNet RSSM inferred 0.700 35/50 1.604 553.03 22.62 484.0
TD-MPC2 Dynamics inferred 0.980 49/50 0.671 513.37 22.84 484.4
PID/LOS controller inferred 0.580 29/50 1.071 300.71 19.32 390.5
No-Flow LOS Controller inferred 0.040 2/50 1.940 449.75 23.56 579.0
Current-Estimator LOS Controller inferred 0.840 42/50 0.678 299.95 18.57 438.4
Oracle-Flow LOS Controller inferred 0.020 1/50 1.878 302.00 18.81 375.0

waypoint_square_twin_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.782 310.81 15.60 328.7
FlowMo shuffled 0.680 34/50 1.464 610.89 26.09 482.0
FlowMo zero 0.680 34/50 1.540 718.58 29.58 527.9
LeWorldModel inferred 0.920 46/50 0.941 561.26 25.63 465.1
PlaNet RSSM inferred 0.620 31/50 1.786 498.65 21.41 436.2
TD-MPC2 Dynamics inferred 0.980 49/50 0.727 450.90 21.67 425.4
PID/LOS controller inferred 0.520 26/50 1.116 298.07 19.11 382.7
No-Flow LOS Controller inferred 0.040 2/50 1.777 388.96 21.97 499.0
Current-Estimator LOS Controller inferred 0.800 40/50 0.806 290.91 18.43 426.8
Oracle-Flow LOS Controller inferred 0.020 1/50 1.698 385.29 19.33 497.0

waypoint_square_twin_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.920 46/50 1.144 338.80 16.29 351.2
FlowMo shuffled 0.560 28/50 1.933 591.62 24.53 464.4
FlowMo zero 0.520 26/50 1.881 662.33 26.94 481.6
LeWorldModel inferred 0.840 42/50 1.250 599.12 26.86 494.7
PlaNet RSSM inferred 0.640 32/50 1.721 521.83 22.21 437.7
TD-MPC2 Dynamics inferred 0.920 46/50 0.937 475.66 21.73 447.2
PID/LOS controller inferred 0.700 35/50 1.208 337.22 20.63 427.2
No-Flow LOS Controller inferred 0.140 7/50 1.674 341.41 21.70 459.7
Current-Estimator LOS Controller inferred 0.860 43/50 1.119 295.43 18.98 440.7
Oracle-Flow LOS Controller inferred 0.060 3/50 1.639 291.97 19.79 366.0

waypoint_square_twin_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.646 342.41 16.14 360.5
FlowMo shuffled 0.560 28/50 1.782 596.53 24.12 483.8
FlowMo zero 0.560 28/50 1.638 709.03 28.93 518.1
LeWorldModel inferred 0.860 43/50 1.220 569.59 24.63 471.1
PlaNet RSSM inferred 0.680 34/50 1.526 514.68 21.28 442.0
TD-MPC2 Dynamics inferred 0.940 47/50 0.888 522.21 22.83 496.5
PID/LOS controller inferred 0.460 23/50 1.042 299.69 18.96 385.9
No-Flow LOS Controller inferred 0.040 2/50 1.830 373.10 22.94 569.0
Current-Estimator LOS Controller inferred 0.780 39/50 0.786 291.58 18.37 434.7
Oracle-Flow LOS Controller inferred 0.020 1/50 1.934 378.60 19.87 509.0

waypoint_zigzag_triangle_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.697 365.88 16.71 360.4
FlowMo shuffled 0.920 46/50 0.666 649.07 25.47 445.9
FlowMo zero 1.000 50/50 0.627 652.71 24.07 407.1
LeWorldModel inferred 1.000 50/50 0.624 569.26 21.77 357.4
PlaNet RSSM inferred 1.000 50/50 0.628 426.37 20.12 381.9
TD-MPC2 Dynamics inferred 0.980 49/50 0.722 503.09 19.99 327.7
PID/LOS controller inferred 1.000 50/50 0.632 293.23 17.91 564.8
No-Flow LOS Controller inferred 1.000 50/50 0.628 257.63 18.23 507.4
Current-Estimator LOS Controller inferred 1.000 50/50 0.625 258.57 16.80 480.9
Oracle-Flow LOS Controller inferred 1.000 50/50 0.626 256.08 17.79 498.8

waypoint_zigzag_triangle_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.637 369.54 16.63 355.1
FlowMo shuffled 0.800 40/50 0.991 661.06 25.42 457.6
FlowMo zero 1.000 50/50 0.626 637.81 23.73 394.5
LeWorldModel inferred 0.920 46/50 0.787 590.49 22.41 368.9
PlaNet RSSM inferred 0.980 49/50 0.789 426.20 19.81 373.7
TD-MPC2 Dynamics inferred 1.000 50/50 0.613 501.40 20.36 335.3
PID/LOS controller inferred 1.000 50/50 0.626 293.46 17.49 544.6
No-Flow LOS Controller inferred 1.000 50/50 0.620 257.53 17.89 493.6
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 257.34 16.92 477.6
Oracle-Flow LOS Controller inferred 1.000 50/50 0.624 256.57 17.78 490.9

waypoint_zigzag_triangle_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.635 366.70 16.67 335.5
FlowMo shuffled 0.920 46/50 0.782 669.98 26.05 463.7
FlowMo zero 1.000 50/50 0.626 675.36 24.82 417.6
LeWorldModel inferred 0.820 41/50 0.973 596.71 22.60 371.4
PlaNet RSSM inferred 0.960 48/50 0.903 413.91 19.71 368.5
TD-MPC2 Dynamics inferred 1.000 50/50 0.612 487.61 19.38 316.4
PID/LOS controller inferred 1.000 50/50 0.629 294.60 17.26 534.0
No-Flow LOS Controller inferred 1.000 50/50 0.622 257.30 17.73 486.4
Current-Estimator LOS Controller inferred 1.000 50/50 0.625 260.43 16.97 476.1
Oracle-Flow LOS Controller inferred 1.000 50/50 0.622 257.30 17.73 486.4

waypoint_zigzag_triangle_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.637 378.80 17.10 359.1
FlowMo shuffled 0.860 43/50 0.761 662.51 26.24 457.4
FlowMo zero 1.000 50/50 0.627 667.28 24.44 405.6
LeWorldModel inferred 0.940 47/50 0.740 584.10 22.44 359.1
PlaNet RSSM inferred 1.000 50/50 0.633 413.36 19.87 359.7
TD-MPC2 Dynamics inferred 1.000 50/50 0.618 481.04 19.79 318.1
PID/LOS controller inferred 1.000 50/50 0.626 296.55 17.52 532.0
No-Flow LOS Controller inferred 1.000 50/50 0.624 255.63 17.92 482.8
Current-Estimator LOS Controller inferred 1.000 50/50 0.632 260.35 17.02 471.4
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 257.63 17.77 480.8

waypoint_zigzag_triangle_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.636 370.00 16.92 368.2
FlowMo shuffled 0.800 40/50 0.956 629.35 24.74 413.9
FlowMo zero 1.000 50/50 0.624 653.54 24.23 393.8
LeWorldModel inferred 0.920 46/50 0.786 577.56 22.29 363.9
PlaNet RSSM inferred 0.980 49/50 0.782 420.27 20.16 371.6
TD-MPC2 Dynamics inferred 1.000 50/50 0.611 490.66 19.89 323.4
PID/LOS controller inferred 1.000 50/50 0.628 285.28 17.56 532.0
No-Flow LOS Controller inferred 1.000 50/50 0.624 249.49 17.97 483.0
Current-Estimator LOS Controller inferred 1.000 50/50 0.623 251.68 16.95 469.3
Oracle-Flow LOS Controller inferred 1.000 50/50 0.625 251.11 17.85 481.8

waypoint_zigzag_triangle_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.689 372.43 16.65 348.2
FlowMo shuffled 0.900 45/50 0.781 644.37 24.83 440.3
FlowMo zero 1.000 50/50 0.626 696.25 24.89 425.3
LeWorldModel inferred 0.880 44/50 0.905 622.90 22.59 381.7
PlaNet RSSM inferred 1.000 50/50 0.632 440.68 20.33 397.3
TD-MPC2 Dynamics inferred 1.000 50/50 0.610 498.23 19.77 335.1
PID/LOS controller inferred 1.000 50/50 0.630 301.73 17.28 557.6
No-Flow LOS Controller inferred 1.000 50/50 0.626 263.50 17.67 503.8
Current-Estimator LOS Controller inferred 1.000 50/50 0.629 266.82 16.90 492.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 263.12 17.59 502.1

waypoint_zigzag_triangle_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.636 362.79 16.50 353.6
FlowMo shuffled 0.920 46/50 0.733 676.61 26.66 473.2
FlowMo zero 1.000 50/50 0.623 671.17 24.90 406.7
LeWorldModel inferred 0.880 44/50 0.889 558.44 21.77 349.3
PlaNet RSSM inferred 1.000 50/50 0.632 438.90 20.62 389.4
TD-MPC2 Dynamics inferred 1.000 50/50 0.611 479.00 19.84 323.5
PID/LOS controller inferred 1.000 50/50 0.632 294.07 17.40 547.1
No-Flow LOS Controller inferred 1.000 50/50 0.628 257.08 17.78 493.7
Current-Estimator LOS Controller inferred 1.000 50/50 0.629 259.71 16.83 480.0
Oracle-Flow LOS Controller inferred 1.000 50/50 0.629 255.90 17.64 490.4

waypoint_zigzag_triangle_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 1.000 50/50 0.634 360.36 16.66 340.9
FlowMo shuffled 0.940 47/50 0.665 682.66 27.48 455.4
FlowMo zero 1.000 50/50 0.622 672.63 25.51 416.8
LeWorldModel inferred 0.900 45/50 0.830 568.62 22.37 357.6
PlaNet RSSM inferred 0.980 49/50 0.775 393.35 19.41 358.5
TD-MPC2 Dynamics inferred 1.000 50/50 0.614 478.78 19.74 313.8
PID/LOS controller inferred 1.000 50/50 0.627 290.20 17.32 527.2
No-Flow LOS Controller inferred 1.000 50/50 0.627 250.91 17.75 477.8
Current-Estimator LOS Controller inferred 1.000 50/50 0.628 251.32 16.78 462.7
Oracle-Flow LOS Controller inferred 1.000 50/50 0.622 251.44 17.70 476.9

waypoint_zigzag_triangle_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.638 372.85 16.57 341.8
FlowMo shuffled 0.880 44/50 0.773 661.58 26.01 462.4
FlowMo zero 0.980 49/50 0.639 655.73 24.46 403.4
LeWorldModel inferred 1.000 50/50 0.626 585.43 22.10 361.7
PlaNet RSSM inferred 1.000 50/50 0.629 414.68 19.50 371.1
TD-MPC2 Dynamics inferred 1.000 50/50 0.616 485.33 19.48 320.3
PID/LOS controller inferred 1.000 50/50 0.631 297.58 17.62 559.2
No-Flow LOS Controller inferred 1.000 50/50 0.626 257.77 18.01 503.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.629 259.25 16.88 484.1
Oracle-Flow LOS Controller inferred 1.000 50/50 0.628 256.14 17.77 498.3

waypoint_zigzag_triangle_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.641 378.76 16.81 358.5
FlowMo shuffled 0.900 45/50 0.795 635.63 24.05 416.8
FlowMo zero 1.000 50/50 0.625 694.41 25.18 431.7
LeWorldModel inferred 0.880 44/50 0.863 589.78 22.07 365.0
PlaNet RSSM inferred 0.980 49/50 0.779 435.24 20.25 383.6
TD-MPC2 Dynamics inferred 1.000 50/50 0.611 494.66 19.61 332.2
PID/LOS controller inferred 1.000 50/50 0.634 302.06 17.33 556.9
No-Flow LOS Controller inferred 1.000 50/50 0.627 263.09 17.74 503.1
Current-Estimator LOS Controller inferred 1.000 50/50 0.627 265.21 16.94 491.2
Oracle-Flow LOS Controller inferred 1.000 50/50 0.627 262.66 17.67 501.8

waypoint_zigzag_twin_double_gyre

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.920 46/50 0.786 445.83 20.09 482.4
FlowMo shuffled 0.740 37/50 1.324 679.53 28.25 548.2
FlowMo zero 0.620 31/50 1.709 802.79 32.27 562.9
LeWorldModel inferred 0.860 43/50 0.980 616.90 25.83 481.3
PlaNet RSSM inferred 0.940 47/50 0.737 610.56 25.72 542.5
TD-MPC2 Dynamics inferred 0.880 44/50 0.973 492.50 23.41 444.2
PID/LOS controller inferred 0.820 41/50 0.840 417.63 27.63 620.6
No-Flow LOS Controller inferred 0.160 8/50 1.770 406.78 29.70 639.1
Current-Estimator LOS Controller inferred 0.860 43/50 0.688 387.65 24.18 571.8
Oracle-Flow LOS Controller inferred 0.260 13/50 1.473 409.15 29.60 630.2

waypoint_zigzag_twin_gradient

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.680 440.64 19.53 476.8
FlowMo shuffled 0.680 34/50 1.126 655.19 28.27 523.6
FlowMo zero 0.740 37/50 1.419 826.28 34.18 581.1
LeWorldModel inferred 0.900 45/50 0.911 637.89 26.89 492.1
PlaNet RSSM inferred 0.760 38/50 1.422 567.48 24.40 501.2
TD-MPC2 Dynamics inferred 0.860 43/50 1.061 476.91 22.47 432.7
PID/LOS controller inferred 0.900 45/50 0.679 434.99 27.57 630.9
No-Flow LOS Controller inferred 0.160 8/50 1.418 410.70 30.62 665.6
Current-Estimator LOS Controller inferred 0.860 43/50 0.669 391.20 24.34 554.7
Oracle-Flow LOS Controller inferred 0.200 10/50 1.387 406.00 29.63 626.6

waypoint_zigzag_twin_noflow

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.686 431.82 19.41 466.0
FlowMo shuffled 0.740 37/50 1.341 678.03 28.74 516.1
FlowMo zero 0.620 31/50 1.569 842.53 33.57 591.2
LeWorldModel inferred 0.880 44/50 0.952 610.92 25.95 475.5
PlaNet RSSM inferred 0.880 44/50 0.964 570.38 24.30 493.2
TD-MPC2 Dynamics inferred 0.920 46/50 0.879 459.15 21.30 404.9
PID/LOS controller inferred 0.920 46/50 0.678 427.72 27.37 629.0
No-Flow LOS Controller inferred 0.280 14/50 1.386 413.61 28.68 643.2
Current-Estimator LOS Controller inferred 0.840 42/50 0.685 385.03 23.81 541.1
Oracle-Flow LOS Controller inferred 0.280 14/50 1.386 413.61 28.68 643.2

waypoint_zigzag_twin_random_fourier

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.686 437.30 19.46 465.6
FlowMo shuffled 0.740 37/50 1.584 680.46 29.19 533.2
FlowMo zero 0.720 36/50 1.419 783.72 32.14 552.3
LeWorldModel inferred 0.920 46/50 0.907 618.01 26.11 490.8
PlaNet RSSM inferred 0.820 41/50 1.197 575.06 24.85 505.5
TD-MPC2 Dynamics inferred 0.880 44/50 1.024 462.50 22.53 423.9
PID/LOS controller inferred 0.860 43/50 0.791 423.45 27.89 625.5
No-Flow LOS Controller inferred 0.280 14/50 1.320 407.97 30.25 638.6
Current-Estimator LOS Controller inferred 0.860 43/50 0.696 385.10 24.59 543.7
Oracle-Flow LOS Controller inferred 0.280 14/50 1.465 411.88 30.16 656.2

waypoint_zigzag_twin_shear

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.691 402.04 18.65 443.2
FlowMo shuffled 0.760 38/50 1.191 681.38 29.91 543.9
FlowMo zero 0.560 28/50 1.761 734.35 30.88 516.9
LeWorldModel inferred 0.860 43/50 0.982 599.40 26.03 482.8
PlaNet RSSM inferred 0.800 40/50 1.368 579.67 25.06 501.0
TD-MPC2 Dynamics inferred 0.860 43/50 1.147 450.85 21.97 414.2
PID/LOS controller inferred 0.780 39/50 0.829 425.40 28.94 610.9
No-Flow LOS Controller inferred 0.180 9/50 1.628 389.70 30.44 570.0
Current-Estimator LOS Controller inferred 0.900 45/50 0.935 375.83 24.22 526.3
Oracle-Flow LOS Controller inferred 0.260 13/50 1.337 400.10 30.53 609.8

waypoint_zigzag_twin_source_sink

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.940 47/50 0.886 457.18 19.61 490.4
FlowMo shuffled 0.720 36/50 1.213 699.21 28.07 549.6
FlowMo zero 0.620 31/50 1.543 794.38 31.55 559.0
LeWorldModel inferred 0.900 45/50 0.826 647.03 26.92 517.4
PlaNet RSSM inferred 0.820 41/50 1.026 605.10 24.70 520.3
TD-MPC2 Dynamics inferred 0.860 43/50 1.004 491.60 21.60 450.1
PID/LOS controller inferred 0.840 42/50 0.732 448.12 27.28 649.0
No-Flow LOS Controller inferred 0.180 9/50 1.520 406.13 28.71 625.1
Current-Estimator LOS Controller inferred 0.760 38/50 0.691 395.60 24.12 554.8
Oracle-Flow LOS Controller inferred 0.200 10/50 1.529 415.61 28.11 621.3

waypoint_zigzag_twin_source_sink_pair

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.980 49/50 0.683 426.05 18.92 463.2
FlowMo shuffled 0.660 33/50 1.560 710.33 29.04 580.0
FlowMo zero 0.680 34/50 1.424 815.96 32.15 571.4
LeWorldModel inferred 0.960 48/50 0.706 614.22 25.84 487.1
PlaNet RSSM inferred 0.840 42/50 0.864 581.77 24.72 509.3
TD-MPC2 Dynamics inferred 0.800 40/50 1.246 452.21 21.70 421.5
PID/LOS controller inferred 0.800 40/50 0.735 443.41 26.74 639.5
No-Flow LOS Controller inferred 0.220 11/50 1.317 435.09 28.62 660.8
Current-Estimator LOS Controller inferred 0.820 41/50 0.679 413.54 24.32 585.9
Oracle-Flow LOS Controller inferred 0.240 12/50 1.262 424.53 28.55 657.1

waypoint_zigzag_twin_turbulent_patch

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.920 46/50 0.697 420.08 19.35 457.5
FlowMo shuffled 0.700 35/50 1.408 684.45 29.85 538.2
FlowMo zero 0.720 36/50 1.387 783.96 32.70 549.4
LeWorldModel inferred 0.920 46/50 0.841 619.63 26.78 488.5
PlaNet RSSM inferred 0.760 38/50 1.582 584.34 25.19 514.3
TD-MPC2 Dynamics inferred 0.840 42/50 1.122 467.69 22.17 426.4
PID/LOS controller inferred 0.900 45/50 0.678 424.39 27.11 614.3
No-Flow LOS Controller inferred 0.300 15/50 1.381 425.51 29.34 650.7
Current-Estimator LOS Controller inferred 0.780 39/50 0.688 377.47 23.72 534.5
Oracle-Flow LOS Controller inferred 0.300 15/50 1.491 421.47 29.03 655.7

waypoint_zigzag_twin_uniform

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.684 438.73 20.05 473.7
FlowMo shuffled 0.700 35/50 1.433 689.87 28.04 545.2
FlowMo zero 0.640 32/50 1.731 774.93 31.88 545.6
LeWorldModel inferred 0.920 46/50 0.783 572.87 24.70 450.4
PlaNet RSSM inferred 0.900 45/50 0.921 567.50 24.42 495.9
TD-MPC2 Dynamics inferred 0.900 45/50 0.904 470.97 22.12 423.7
PID/LOS controller inferred 0.920 46/50 0.779 430.42 27.21 623.9
No-Flow LOS Controller inferred 0.280 14/50 1.149 399.15 29.32 602.4
Current-Estimator LOS Controller inferred 0.860 43/50 0.674 386.36 24.24 560.0
Oracle-Flow LOS Controller inferred 0.340 17/50 1.109 397.63 29.20 622.6

waypoint_zigzag_twin_vortex_center

Method Context Success Successes Final Dist Energy (Success) Path Len (Success) Steps (Success)
FlowMo inferred 0.960 48/50 0.739 450.92 19.66 484.7
FlowMo shuffled 0.720 36/50 1.088 711.42 28.54 552.1
FlowMo zero 0.720 36/50 1.254 861.98 33.93 604.9
LeWorldModel inferred 0.740 37/50 1.356 666.10 26.60 526.2
PlaNet RSSM inferred 0.800 40/50 1.252 627.72 25.60 537.2
TD-MPC2 Dynamics inferred 0.780 39/50 1.315 500.39 21.95 450.3
PID/LOS controller inferred 0.800 40/50 1.008 444.32 27.25 643.3
No-Flow LOS Controller inferred 0.240 12/50 1.387 416.93 29.62 669.5
Current-Estimator LOS Controller inferred 0.740 37/50 0.796 404.83 23.98 574.8
Oracle-Flow LOS Controller inferred 0.280 14/50 1.349 423.87 29.74 657.1

Artifact Locations

  • Prediction JSON: experiments/reports/paper_prediction.json
  • FlowMo latent probe JSON: experiments/reports/paper_flowmo_latent_probes.json
  • Planning JSON files: experiments/reports/paper_planning/*.json
  • GIFs: experiments/reports/paper_planning/gifs/*.gif