Table 13 Obstacle density sensitivity analysis (20 runs per condition).

From: A hybrid APF-DQN framework with transformer-based current prediction for USV path planning in dynamic ocean environments

Density

Method

Success (%)

Path Len. (m)

Steps

Energy

15% (Sparse)

E-APF

100.0

263.11

46.00

50.46

DQN

75.0

\(568.62 \pm 185.66\)

\(114.07 \pm 42.88\)

\(184.44 \pm 81.29\)

APF-DQN

100.0

\(259.91 \pm 5.00\)

\(42.75 \pm 5.06\)

\(43.14 \pm 7.89\)

25% (Standard)

E-APF

100.0

270.09

47.00

53.67

DQN

100.0

\(294.69 \pm 20.86\)

\(50.85 \pm 6.17\)

\(76.04 \pm 13.30\)

APF-DQN

100.0

\(262.85 \pm 4.18\)

\(41.30 \pm 0.84\)

\(49.55 \pm 2.05\)

35% (Dense)

E-APF

80.0

\(395.16 \pm 180.51\)

\(93.38 \pm 42.45\)

\(126.39 \pm 82.07\)

DQN

20.0

\(451.09 \pm 53.16\)

\(141.00 \pm 29.47\)

\(164.02 \pm 12.37\)

APF-DQN

100.0

\(275.51 \pm 8.94\)

\(54.50 \pm 8.67\)

\(70.88 \pm 12.79\)