Results with Obstacles

Companion page: obstacle-avoidance experiments for all environments. See the main page for the full results.

Dubins Car

Diffusion

STLPY

Note: TtR = mean time-to-reach (steps).

Results with obstacles for the Dubins Car environment (Heterogeneous specs; $N\in\{8,16,32\}$). D-MA: Diffusion-based Multi-Agent (Ours), D-SA: Diffusion-based Single-Agent, STLPY: MILP planner. Best results are in bold.
Spec N Planning Time (s) (↓) Success Rate (↑) TtR (↓)
D-MAD-SASTLPY D-MAD-SASTLPY D-MAD-SASTLPY
Branch 8 0.270.2211.49 95.0090.4283.75 1334.661234.21670.08
16 0.840.3122.30 90.6282.5078.12 1341.261327.891045.21
32 2.260.4644.12 87.1955.0066.88 1753.571620.55998.40
Cover 8 0.830.429.12 90.0087.0885.00 1599.751246.25854.22
16 1.380.5818.37 90.0079.3880.62 1659.241382.991141.86
32 4.140.8137.35 82.5055.4262.81 1591.701638.021097.58
Loop 8 0.841.8060.62 87.9283.3372.50 2625.002167.891795.36
16 1.422.55126.32 85.0080.8373.75 2511.692386.081880.43
32 2.192.98240.85 71.0461.9843.75 2858.892837.732101.98
Seq. 8 0.430.324.59 87.9285.4276.25 1676.991727.931295.20
16 1.870.429.08 88.7575.6268.75 1878.021747.221529.92
32 4.750.6818.86 71.2542.1941.88 1843.992022.891799.16

Double Integrator

Diffusion

STLPY

Note: TtR = mean time-to-reach (steps).

Results with obstacles for the Double Integrator environment (Heterogeneous specs; $N\in\{8,16,32\}$). D-MA: Diffusion-based Multi-Agent (Ours), STLPY: MILP planner. Best results are in bold.
Spec N Planning Time (s) (↓) Success Rate (↑) TtR (↓)
D-MASTLPY D-MASTLPY D-MASTLPY
Branch 8 0.579.91 97.9285.00 1466.511136.44
16 0.9319.32 98.7588.12 1686.291203.41
32 2.1739.47 97.6083.75 1940.351458.93
Cover 8 1.568.31 98.7585.00 1389.641177.51
16 2.2816.55 96.8887.50 1607.361268.25
32 4.4332.81 87.8176.88 1908.441343.73
Loop 8 3.9549.58 92.9287.50 2315.571871.11
16 7.52100.53 93.5488.12 2379.242097.43
32 14.43196.92 90.2183.75 3124.323058.88
Seq. 8 2.033.98 96.2582.50 1580.931234.13
16 2.547.97 96.0483.75 1834.801780.33
32 4.4215.81 85.7366.25 2324.781850.46

Single Integrator

Diffusion

STLPY

Note: TtR = mean time-to-reach (steps).

Results with obstacles for the Single Integrator environment (Heterogeneous specs; $N\in\{8,16,32\}$). D-MA: Diffusion-based Multi-Agent (Ours), D-SA: Diffusion-based Single-Agent, STLPY: MILP planner. Best results are in bold.
Spec N Planning Time (s) (↓) Success Rate (↑) TtR (↓)
D-MAD-SASTLPY D-MAD-SASTLPY D-MAD-SASTLPY
Branch 8 0.540.279.93 94.1789.5881.25 1234.491489.94665.80
16 0.730.3019.29 93.5490.0085.62 1323.921549.64709.70
32 1.070.3138.62 92.8191.6787.19 1778.981677.16734.18
Cover 8 0.420.318.18 92.9285.8381.25 1430.521479.96853.77
16 0.620.4316.45 92.2988.1285.62 1529.051540.77887.44
32 1.090.4732.89 92.8189.9088.12 1555.411448.32947.44
Loop 8 3.771.8652.86 90.8381.2578.75 1904.901856.041233.15
16 7.312.55105.73 90.8386.6781.88 1886.951952.121119.33
32 12.763.92218.89 90.7387.1984.38 2206.912154.671423.70
Seq. 8 1.250.454.11 91.6789.5881.25 1573.661715.81947.60
16 1.810.508.12 93.3388.7583.12 1721.081824.431006.02
32 3.350.6416.39 91.7787.5080.94 2243.081988.051387.37