Table 4 Reward function design parameters.

From: Deep reinforcement learning model for Multi-Ship collision avoidance decision making design implementation and performance analysis

Parameter

Value range

Description

\(\:\alpha\:\)

[0, 1]

Weight for collision avoidance reward

\(\:\beta\:\)

[0, 1]

Weight for navigation efficiency reward

\(\:\gamma\:\)

[0, 1]

Weight for COLREGs compliance reward

\(\:{d}_{safe}\)

> 0

Minimum safe distance between ships

\(\:{d}_{max}\)

> \(\:{d}_{safe}\)

Maximum distance threshold for collision avoidance reward