Lidar-Based Distance Map
NVIDIA Docs Hub  NVIDIA Isaac AMR
  Isaac AMR  Lidar-Based Distance Map

Overview

The Lidar-Based Distance Map is a grid-based representation of the immediate environment surrounding the robot. It can take inputs from multiple Lidar sensors, where each Lidar sensor follows an independent firing pattern and can be arbitrarily oriented on the robot. By combining the data from these Lidar sensors, it generates a unified distance map that is used by the planning stack.

lidar_distance_map_arch.png

  1. Upon receiving a Lidar sensor reading, the LidarToEvidenceMap module represents the range values using the bird’s-eye evidence grid map. In this map, grid cells before the hit point are marked as “free”, while those beyond the hit point are marked as “uncertain”. The Lidar lattice specifies the center of the sensor on the grid map and aids in visualizing the sensor range values in relation to its own lattice.

  2. Once a sensor evidence grid is generated, it is fused with the unified grid computed in the previous time step within the EvidenceMapFusion codelet. The previously computed fused map is transformed based on the relative movement of the robot between time steps. Additionally, the sensor map is transformed based on the alignment of the sensor with respect to the robot, as specified in the robot model. Both transformed maps are then stacked, and their corresponding evidence grid values are fused together, interpolating where necessary.

  3. The EvidenceMapInpaint codelet allows the user to manually assign evidence-mass values to specific areas in the evidence grid map. This is particularly useful in scenarios where the sensor configuration has blind spots or areas that require special attention.

  4. The EvidenceToBinaryMap module converts the evidence map into a binary map by applying parameterized thresholds to differentiate between free and occupied classes.

  5. Finally, the BinaryToDistanceMap codelet converts the binary map into a distance map, which is subsequently used by the planner for further processing.

Evidence Grid Map

To fuse the data from different Lidar sensors and generate a unified distance map, an intermediate grid-based representation called the “evidence grid map” is utilized. The Dempster–Shafer Theory is employed to construct an occupancy grid around the robot, where each grid cell contains three floating-point values: the belief mass that the cell is free, the belief mass that the cell is occupied, and the leftover mass that is assigned as uncertain. Each sensor message is parsed into an evidence grid map, and these individual grid maps are fused together to construct a unified evidence grid map of the environment.

Below is an illustration of a representative factory environment and its corresponding unified evidence grid map:

egm_factory_example.png

API Documentation

NvIsaacEgmFusionExtension

Extension containing EGM Fusion related components.

nvidia::isaac::BinaryToDistanceMap

Converts binary maps to distance maps

Parameters:

name

Allocator
lookup_name

allocator
description

Memory allocator used for tensor
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Binary Map
lookup_name

binary_map
description

Receiver for binary map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Distance Map
lookup_name

distance_map
description

Transmitter for generated distance map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Binary Map Size
lookup_name

binary_map_size
description

Size of binary map tensor (rows, cols)
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Compute Inside Distance
lookup_name

compute_inside_distance
description

Whether to compute the distances inside obstacles (negative distances)
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

False
name

Maximum Distance
lookup_name

maximum_distance
description

Maximum distance to include in the generated distance map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT64
default

20.0
name

Number of Threads Per Block
lookup_name

number_threads_per_block
description

The number of threads per block used in the GPU when computing the distance map.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

32
name

CUDA Stream
lookup_name

cuda_stream
description

CUDA Stream Holder to provide the stream to put load on.
flags

GXF_PARAMETER_FLAGS_OPTIONAL
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Inpute is Device Tensor
lookup_name

input_tensor_device
description

Whether binary_map input tensor is on device memory
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

False

nvidia::isaac::BinaryMapInpaint

Paints rectangles on the binary map.

Parameters:

name

Map To Paint
lookup_name

map_to_paint
description

Name of the channel for incoming binary map that needs to be painted
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Painted Binary Map
lookup_name

painted_binary_map
description

Name of the outgoing binary map channel with painted rectangles
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Paint mass
lookup_name

paint_mass
description

The mass value that will be used to paint every cell within the rectangular region
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_UINT8
default

N/A
name

Rectangle
lookup_name

rectangle
description

Rectangular region to paint in the binary grid, with the format[[row_start, row_end], [col_start, col_end]] in grid cells where grid starts with 0, 0 atthe top left
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

CUDA Stream
lookup_name

cuda_stream
description

CUDA Stream Holder to provide the stream to put load on.
flags

GXF_PARAMETER_FLAGS_OPTIONAL
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Asynchronous Mode
lookup_name

async_mode
description

Experimental: Exit after input buffers are consumed not waiting for output buffer if true.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

False

nvidia::isaac::ClassicalMapFusion

Fuses evidence maps based on classical fusion rules

Parameters:

name

Allocator
lookup_name

allocator
description

Memory allocator used for tensor
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Sensor Evidence Grid
lookup_name

sensor_evidence_grid
description

Name of input sensor map channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Fused Map Lattice
lookup_name

fused_map_lattice
description

Name of the input channel for defining the fused map lattice
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Local Map Lattice
lookup_name

local_map_lattice
description

Name of the input channel for defining the local map lattice
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Fused Evidence Grid
lookup_name

fused_evidence_grid
description

Name of output fused sensor map channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Local Map Dimensions
lookup_name

local_map_dimensions
description

Number of rows and columns in the output evidence grid
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Fused Map Dimensions
lookup_name

fused_map_dimensions
description

Number of rows and columns in the fused evidence grid
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Maximum Ratio of Fused Map Delta
lookup_name

maximum_fused_map_delta
description

Ratio of maximum change of fused map center between ticks over the fused map dimensions.Fusion reports failure if the fused map center change exceeds this ratio.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT64
default

0.5
name

Odometry Frame
lookup_name

odometry_frame
description

PoseTreeFrame of the odometry frame in the pose tree
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Base Frame
lookup_name

base_frame
description

PoseTreeFrame of the base frame of the robot in the pose tree
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

ATLAS
lookup_name

atlas
description

Link to the ATLAS backend
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Fusion Rule
lookup_name

fusion_rule
description

ID of the fusion operator to use. 0 for the Dempster Shafer fusion rule, 1 for the JosangAveraging rule, 2 for the Josang Cumulative rule, 3 for the PCR6 CPU rule and 4 for the PCR6cuda based rule
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

4
name

Cuda Stream
lookup_name

cuda_stream
description

Instance of nvidia::isaac::CudaStream to grab CUDA stream from.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::EvidenceMapInpaint

Paints rectangles on the evidence map.

Parameters:

name

Map To Paint
lookup_name

map_to_paint
description

Name of the channel for incoming evidence map that needs to be painted
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Painted Evidence Map
lookup_name

painted_evidence_map
description

Name of the outgoing evidence map channel with painted rectangles
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Rectangles Frame
lookup_name

rectangles_frame
description

The PoseTreeFrame of the rectangular region to paint
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Paint mass
lookup_name

paint_mass
description

The mass values that will be used to paint every cell within the rectangular region
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Rectangles
lookup_name

rectangles
description

Set of rectangular regions to paint in the evidence grid, each one with the format[[x_min,x_max], [y_min,y_max]] in meters
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

ATLAS
lookup_name

atlas
description

Link to ATLAS backend
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::EvidenceMapVisualizer

Visualizes an evidence grid map on Sight

Parameters:

name

Transmitter
lookup_name

transmitter
description

Transmit the evidence grid to the Sight window
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Evidence Map
lookup_name

evidence_map
description

Evidence grid map that needs to be visualized on Sight
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Uncertainity Color
lookup_name

uncertainity_color
description

RGB color for the uncertain region in an evidence grid
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

ATLAS
lookup_name

atlas
description

Link to the ATLAS backend
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Allocator
lookup_name

allocator
description

Allocator for allocating image space.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::EvidenceMapWriter

Writes an evidence grid map to a RGB image

Parameters:

name

ATLAS
lookup_name

atlas
description

Link to the ATLAS frontend
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Allocator
lookup_name

allocator
description

Allocator for allocating image space.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Output file base path
lookup_name

output_file_base_path
description

Base path for the output file
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Evidence Map Receiver
lookup_name

rx_evidence_map
description

Evidence grid map receiver channel.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Color Scheme
lookup_name

color_scheme
description

Scheme option to colorize each class of EGM (Free, Occupied, Unknown) in RGB images.0: Red-Green-Black (default);1: MixBelief (White-Black-UncertaintyColor);2: MaxBelief (White-Black-UncertaintyColor)
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

0
name

Uncertainty Color
lookup_name

uncertainty_color
description

RGB color for the uncertain region in an evidence grid.Used only in color scheme 2&3.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A

nvidia::isaac::EvidenceBinarizer

Converts evidence map into a binary map with different strategies.

Parameters:

nvidia::isaac::EvidenceBinarizerOccupancy

Converts evidence map into a occupancy map based on the relative evidence between free and occupied.

Parameters:

name

Free Threshold
lookup_name

free_threshold
description

Minimum difference between the free and occupied mass of a cell for it to be labelled free
flags

GXF_PARAMETER_FLAGS_DYNAMIC
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.10000000149011612
name

Occupied Threshold
lookup_name

occupied_threshold
description

Minimum difference between the occupied and free mass of a cell for it to be labelled occupied
flags

GXF_PARAMETER_FLAGS_DYNAMIC
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.10000000149011612
name

Optimistic Flag
lookup_name

is_optimistic
description

If enabled, uncertain cells will be treated as free, otherwise they are considered to beblocked
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

False

nvidia::isaac::EvidenceBinarizerUncertain

Converts evidence map into a occupancy map by checking if the grids are uncertain

Parameters:

name

Occluded Threshold
lookup_name

occluded_threshold
description

Maximum threshold to decide if a grid is considered as not occluded

flags

GXF_PARAMETER_FLAGS_DYNAMIC
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.10000000149011612
name

Difference Threshold
lookup_name

difference_threshold
description

Maximum difference between the occupied and free mass of a cell to be labelled as uncertain
flags

GXF_PARAMETER_FLAGS_DYNAMIC
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.10000000149011612

nvidia::isaac::EvidenceToBinaryMap

Converts evidence maps to binary maps.

Parameters:

name

Host memory allocator
lookup_name

allocator
description

The memory allocator used for host memory allocation
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Evidence Map
lookup_name

evidence
description

Name of the incoming ‘evidence map’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Binary Map
lookup_name

binary_map
description

Name of the outgoing ‘binary map’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Evidence Binarizer
lookup_name

evidence_binarizer
description

Components that convert evidence into binary flag.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::FlatscanToEvidenceMap

Converts flatscans to evidence maps, according to a lattice.

Parameters:

name

Host memory allocator
lookup_name

allocator
description

The memory allocator used for host memory allocation
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Flatscan
lookup_name

flatscan
description

Name of the incoming ‘flatscan’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Lattice
lookup_name

lattice
description

Name of the incoming ‘lattice’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Evidence Map
lookup_name

evidence_map
description

Name of the outgoing ‘evidence_map’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Wall thickness
lookup_name

wall_thickness
description

The assumed thickness of the walls.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT64
default

N/A
name

ATLAS
lookup_name

atlas
description

Link to the ATLAS backend
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Rows
lookup_name

rows
description

Rows of evidence grid
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A
name

Columns
lookup_name

columns
description

Columns in evidence grid
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A

nvidia::isaac::FusionAssociation

Recursively fuses sensor maps over multiple timesteps.

Parameters:

name

Pool
lookup_name

pool
description

Memory allocator for the output messages.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Current Observation
lookup_name

current_observation
description

Input channel for the current sensor observation map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tx Fused Map
lookup_name

tx_fused_map
description

Tensor map after fusing the sensor observation map with the workmap
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Input Tensor Name
lookup_name

input_tensor_name
description

The name of the tensor in the input observation map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Output Tensor Name
lookup_name

output_tensor_name
description

The name of the tensor in the output fused map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Atlas Frontend
lookup_name

atlas
description

Atlas Frontend to access pose tree.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Odometry Frame
lookup_name

odom_frame
description

A static reference frame for computing robot pose delta between different timestamps.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Robot Pose Frame
lookup_name

robot_frame
description

The robot’s pose frame. USS sensor poses are fixed with respect to robot frame.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Work map lattice
lookup_name

work_map_lattice
description

Name of the input channel for defining the work map lattice
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Sensor map lattice
lookup_name

sensor_map_lattice
description

Name of the input channel for defining the input sensor map lattice
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Visible Decay Rate
lookup_name

visible_decay_rate
description

The discount factor applied to those sections of the workmap which are within the field ofview of the sensors before the observation map is fused with it
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

1.0
name

Non Visible Decay Rate
lookup_name

non_visible_decay_rate
description

The discount factor applied to those sections of the workmap which are outside the field ofview of the sensors before the observation map is fused with it
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

1.0
name

Cuda Stream
lookup_name

cuda_stream
description

Instance of nvidia::isaac::CudaStream to grab CUDA stream from.
flags

GXF_PARAMETER_FLAGS_OPTIONAL
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::LatentUnprojection

Computes the unprojected latent grid map from Ultrasonic prefusion.

Parameters:

name

Pool
lookup_name

pool
description

Memory allocator for the output messages.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Latent Vector
lookup_name

latent_vector
description

Input channel for the latent ultrasonic device tensor.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Intrinsics
lookup_name

intrinsics
description

Input channel for the ultrasonic intrinsics device tensor.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Sensor Pose
lookup_name

sensor_pose
description

Input channel for the sensor pose device tensor.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Output
lookup_name

output
description

Output channel for the unprojected map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Unprojected Map Dimensions
lookup_name

unprojected_map_dimensions
description

Dimensions of the unprojected map.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Grid Centers
lookup_name

grid_centers
description

Pose of the center of the grid with respect to the robot in the format of x, y, angle.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Grid Config
lookup_name

grid_config
description

Specification of the grid containing the length of the grid in meters along the x and ydirections followed by the number of grid cells in both directions.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Latent unprojected map name
lookup_name

latent_map_name
description

The name of the unprojected latent USS map.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Cuda Stream
lookup_name

cuda_stream
description

Instance of nvidia::isaac::CudaStream to grab CUDA stream from.
flags

GXF_PARAMETER_FLAGS_OPTIONAL
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::LinearEmbeddingGenerator

Computes a linear positional embedding tensor

Parameters:

name

Pool
lookup_name

pool
description

Device memory allocator for the output messages.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Output
lookup_name

output
description

Output channel for the embedding map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Embedding Length
lookup_name

embedding_length
description

Length of the embeddings
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A
name

Rows
lookup_name

rows
description

Number of rows in the embedding
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A
name

Batch Size
lookup_name

batch_size
description

Batch size for the embedding
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A
name

Tensor Name
lookup_name

tensor_name
description

The name of the tensor for the output tensor
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A

nvidia::isaac::PrecomputedTensorLoader

Loads precomputed Pytorch tensor data from a binary file

Parameters:

name

Allocator
lookup_name

allocator
description

Allocator for allocating tensor space.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

tensor Shapes
lookup_name

tensor_shapes
description

List of integers defining tensor dimensions in outgoing message.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A
name

File Path
lookup_name

file_path
description

Path to the binary file storing torch tensors.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Storage Type
lookup_name

storage_type
description

Memory storage type for outgoing tensor message - host (0), device (1), or system (2).
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

1
name

Tx Tensor
lookup_name

tx_tensor
description

Transmitter channel name of outgoing message.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tensor Name
lookup_name

tensor_name
description

Name of tensor in the outgoing message.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A

nvidia::isaac::RadialEmbeddingGenerator

Computes a radial positional embedding tensor

Parameters:

name

Pool
lookup_name

pool
description

Device memory allocator for the output messages.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Output
lookup_name

output
description

Output channel for the embedding map
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Grid Dimension
lookup_name

grid_dimension
description

Number of rows and columns in the embedding
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Resolution
lookup_name

resolution
description

Cell size along x and y dimensions
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Robot_T_Grid
lookup_name

robot_T_grid
description

Pose of the ego with respect to the grid
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Batch Size
lookup_name

batch_size
description

Batch size for the embedding
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

N/A
name

Tensor Name
lookup_name

tensor_name
description

The name of the tensor for the output tensor
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A

nvidia::isaac::UssBatchEncoder

Accumulates uss envelope messages and encodes the envelopes, poses and intrinsics as tensors for EGM DNN

Parameters:

name

Trigger Receiver
lookup_name

trigger
description

Message with a Timestamp component used as the target time.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Envelopes Receiver
lookup_name

uss_envelopes
description

USS envelope receiver.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Envelopes Transmitter
lookup_name

tx_envelopes
description

Output tensors containing the accumulated uss envelopes.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Poses Transmitter
lookup_name

tx_poses
description

Output tensors containing the accumulated uss poses.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Intrinsics Transmitter
lookup_name

tx_intrinsics
description

Output tensors containing the accumulated uss intrinsics.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Robot Pose Transmitter
lookup_name

tx_robot_T_odom
description

Output tensors containing the robot pose.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Publish Partial Tensors Flag
lookup_name

publish_partial_tensors
description

If true, all output channels publish their buffers at every tick (even if partially empty)
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

True
name

Atlas Frontend
lookup_name

atlas
description

Atlas Frontend to access pose tree.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Static Pose Frame
lookup_name

static_frame
description

A static reference frame for computing robot pose delta between different timestamps.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Robot Pose Frame
lookup_name

robot_frame
description

The robot’s pose frame. USS sensor poses are fixed with respect to robot frame.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Pose Tree frames for USS Sensors
lookup_name

uss_sensor_frames
description

The names for all ultrasonic sensors on the pose tree where they are supposed to be.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Host Memory allocator
lookup_name

allocator
description

Handle to the memory allocator pool used for data generation.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Ultrasonic Sensors Adjacency List
lookup_name

adjacency
description

List of ultrasonic sensor index pairs to consider adjacent and mutually visiblefor cross envelopes. For example [[0, 1], [1, 2], [3, 4]] means sensor 0 & 1, 1 & 2, 3 & 4 are visible to each other, while 2 & 3 are not.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_CUSTOM
default

N/A
name

Output Number of Envelopes
lookup_name

batch_envelopes_count
description

Number of envelopes in one output message. If less is available in the message queue, the corresponding row is set to zero
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

32
name

Output Number of Bins
lookup_name

batch_bins_count
description

Number of bins in the output envelopes tensor. The input envelop is either truncated/zero-padded at the end when copied to the output envelopes if it has more/less bins than batch_bins_count.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

310
name

Threshold for envelope to target time
lookup_name

latency_threshold
description

Warning is logged if an envelope’s acqtime time to target time exceeds this threshold.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT64
default

0.3

nvidia::isaac::TensorSplit

Splits incoming tensor into multiple sub-tensors of equal size along the first dimension.

Parameters:

name

Allocator
lookup_name

allocator
description

Handle to the memory allocator pool used for data generation.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Rx composed tensor
lookup_name

rx_composed_tensor
description

Channel name of the incoming composed tensor.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tensor Names
lookup_name

tensor_names
description

List of tensor names in outgoing messages.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Tx sliced tensors
lookup_name

tx_sliced_tensors
description

List of channel names of the outgoing sliced tensors
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::TensorStack

Concatenates a sequence of incoming tensors along a new dimension at dimension 0.

Parameters:

name

Allocator
lookup_name

allocator
description

Handle to the memory allocator pool used for data generation.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Rx tensors
lookup_name

rx_tensors
description

List of channel names of the incoming tensors.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tx concatenated tensors
lookup_name

tx_concatenated_tensors
description

Channel name of the outgoing composed tensor.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tensor Name
lookup_name

tensor_name
description

Name of the composed tensor in the outgoing message.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

tensor

nvidia::isaac::TensorToBinaryMap

Converts converts plain tensors to cuda binary map, according to a lattice.

Parameters:

name

Tensor
lookup_name

tensor
description

Name of the incoming ‘tensor’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Lattice
lookup_name

lattice
description

Name of the incoming ‘lattice’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Binary Map
lookup_name

binary_map
description

Name of the outgoing ‘binary_map’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tensor Name
lookup_name

tensor_name
description

The name of the tensor to be extracted in the incoming message.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Rx has explicit Batch Size
lookup_name

rx_has_explicit_batch
description

True if the input tensor is expected to contain the batch dimension
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

True
name

CUDA Stream
lookup_name

cuda_stream
description

CUDA Stream Holder to provide the stream to put load on.
flags

GXF_PARAMETER_FLAGS_OPTIONAL
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

CUDA memory allocator
lookup_name

cuda_allocator
description

The memory allocator used for CUDA memory allocation
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Free Threshold
lookup_name

free_threshold
description

Minimum difference between the free and occupied mass of a cell for it to be labelled free
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.10000000149011612
name

Occupied Threshold
lookup_name

occupied_threshold
description

Minimum difference between the occupied and free mass of a cell for it to be labelled occupied
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.10000000149011612
name

Optimistic Flag
lookup_name

is_optimistic
description

If enabled, uncertain cells will be treated as free, otherwise they are considered to beblocked
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

False
name

Input Mode
lookup_name

input_mode
description

If mode=1, a 3-channel occupancy EGM is expected as input. If mode=2, a 2-channel height EGMis expected
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_INT32
default

1
name

HEGM Height Threshold
lookup_name

hegm_height_threshold
description

Maximum height in car coordinates after which a cell is marked as occupied
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.05000000074505806
name

HEGM Uncertainity Threshold
lookup_name

hegm_uncertainity_threshold
description

Uncertainity value in HEGM after which occupancy is determined by the is_optimistic flag
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.20000000298023224

nvidia::isaac::TensorToEvidenceMap

Converts converts plain tensors to evidence maps, according to a lattice.

Parameters:

name

Host memory allocator
lookup_name

allocator
description

The memory allocator used for host memory allocation
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tensor
lookup_name

tensor
description

Name of the incoming ‘tensor’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Lattice
lookup_name

lattice
description

Name of the incoming ‘lattice’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Evidence Map
lookup_name

evidence_map
description

Name of the outgoing ‘evidence_map’ channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Tensor Name
lookup_name

tensor_name
description

The name of the tensor to be extracted in the incoming message.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_STRING
default

N/A
name

Rx has explicit Batch Size
lookup_name

rx_has_explicit_batch
description

True if the input tensor is expected to contain the batch dimension
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_BOOL
default

True
name

CUDA Stream
lookup_name

cuda_stream
description

CUDA Stream Holder to provide the stream to put load on.
flags

GXF_PARAMETER_FLAGS_OPTIONAL
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

CUDA memory allocator
lookup_name

cuda_allocator
description

The memory allocator used for CUDA memory allocation
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

nvidia::isaac::VegmSynchronization

Publish matched camera messages and egm messages based on closest timestamp.

Parameters:

name

Egm
lookup_name

tx_egm
description

Transmitter channel egm name
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Camera left
lookup_name

tx_camera_left
description

Transmitter channel left camera name
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Camera right
lookup_name

tx_camera_right
description

Transmitter channel right camera name
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Receiver Odometry
lookup_name

rx_odometry
description

Receiver odometry channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Transmitter Odometry
lookup_name

tx_odometry
description

Transmitter odometry channel
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Egm and camera time difference
lookup_name

egm_time_difference
description

Max difference between egm acqtime and camera acqtime to match
flags

GXF_PARAMETER_FLAGS_DYNAMIC
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.05000000074505806
name

Odometry time difference
lookup_name

odometry_time_difference
description

Max difference between egm acqtime and odometry acqtime to match
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.029999999329447746
name

Stereo camera time difference
lookup_name

camera_time_difference
description

Max difference between cam left acqtime and cam right acqtime to match
flags

GXF_PARAMETER_FLAGS_DYNAMIC
type

GXF_PARAMETER_TYPE_FLOAT32
default

0.019999999552965164

nvidia::isaac::EgmFlatsimMetrics

Compute the metrics for EGM against the ground-truth from Flatsim

Parameters:

name

Atlas
lookup_name

atlas
description

Link to Atlas.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Occupancy Map Frame
lookup_name

omap_frame
description

The occupancy map frame.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

EGM Distance Map
lookup_name

rx_egm_distance_map
description

Input for the distance map generated from EGM.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Metric Precision
lookup_name

metric_precision
description

Metric for storing the EGM precision.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A
name

Metric Recall
lookup_name

metric_recall
description

Metric for storing the EGM recall.
flags

GXF_PARAMETER_FLAGS_NONE
type

GXF_PARAMETER_TYPE_HANDLE
default

N/A

© Copyright 2018-2023, NVIDIA Corporation. Last updated on Oct 23, 2023.