navrl_terrain_cfg

NavRL-style discrete obstacles terrain with probability-weighted height intervals.

Extends IsaacLab's height-field terrain to support obstacle_height_mode="range", where obstacle heights are sampled from probability-weighted intervals defined by obstacle_height_range (N+1 boundary values) and obstacle_height_probability (N probabilities summing to 1).

Classes:

Name	Description
HfDiscreteObstaclesNavRLTerrainCfg	Configuration for discrete obstacles terrain with NavRL-style "range" height mode.

Functions:

Name	Description
discrete_obstacles_terrain	Generate a terrain with randomly placed discrete obstacles (pillars).

HfDiscreteObstaclesNavRLTerrainCfg

Bases: HfTerrainBaseCfg

Configuration for discrete obstacles terrain with NavRL-style "range" height mode.

Inherits all base height-field parameters from :class:HfTerrainBaseCfg.

The obstacle_height_mode field selects how obstacle heights are sampled:

• "choice": randomly pick from [-h, -h/2, h/2, h] (IsaacLab default).
• "fixed": always use positive h.
• "range": select an interval from obstacle_height_range based on obstacle_height_probability, then sample uniformly within that interval. In this mode, obstacle_height_range must be a list of N+1 boundary values and obstacle_height_probability a list of N probabilities summing to 1.

Attributes:

Name	Type	Description
obstacle_height_mode	str	The mode to use for obstacle height. Defaults to "range".
obstacle_width_range	tuple[float, float]	The minimum and maximum width of the obstacles (in m).
obstacle_height_range	list	Obstacle height specification.
obstacle_height_probability	list	Probability for each interval in "range" mode. Must sum to 1 and have
num_obstacles	int	The number of obstacles to generate.
platform_width	float	The width of the square platform at the center of the terrain. Defaults to 0.0.

obstacle_height_mode class-attribute instance-attribute

obstacle_height_mode: str = 'range'

The mode to use for obstacle height. Defaults to "range".

Supported modes: "choice", "fixed", "range".

obstacle_width_range class-attribute instance-attribute

obstacle_width_range: tuple[float, float] = MISSING

The minimum and maximum width of the obstacles (in m).

obstacle_height_range class-attribute instance-attribute

obstacle_height_range: list = MISSING

Obstacle height specification.

• For "choice"/"fixed" mode: a 2-element [min, max] in metres.
• For "range" mode: a list of N+1 boundary values (e.g. [1.0, 1.5, 2.0, 4.0, 6.0]).

obstacle_height_probability class-attribute instance-attribute

obstacle_height_probability: list = MISSING

Probability for each interval in "range" mode. Must sum to 1 and have length len(obstacle_height_range) - 1. Ignored for other modes.

num_obstacles class-attribute instance-attribute

num_obstacles: int = MISSING

The number of obstacles to generate.

platform_width class-attribute instance-attribute

platform_width: float = 0.0

The width of the square platform at the center of the terrain. Defaults to 0.0.

discrete_obstacles_terrain

discrete_obstacles_terrain(difficulty: float, cfg: HfDiscreteObstaclesNavRLTerrainCfg) -> np.ndarray

Generate a terrain with randomly placed discrete obstacles (pillars).

Supports three obstacle height modes:

• "choice": randomly pick from [-h, -h/2, h/2, h] (IsaacLab default).
• "fixed": always use positive h.
• "range" (NavRL): select an interval from obstacle_height_range based on obstacle_height_probability, then sample height uniformly within that interval.

Parameters:

Name	Type	Description	Default
difficulty	float	Terrain difficulty in [0, 1].	required
cfg	HfDiscreteObstaclesNavRLTerrainCfg	Terrain configuration.	required

Returns:

Type	Description
ndarray	Height field as a 2D int16 numpy array (width_pixels, length_pixels).