This project is a marker publisher called xela_taxel_viz_ahv4 that visualizes Allegrohand sensor data and publishes it to a marker topic.
I plan to develop a visualization app that subscribes to the sensor data topic provided by xela_server2_ah and publishes visualized markers. Using the force values ​​(taxel values ​​if unavailable) and sensor model (frame_id) information from the sensor data, I will convert the x, y, and z values ​​into markers and visualize them on a grid-based sensor model. As an additional feature, I will link it to the xela_model information to display the corresponding model values ​​as markers at each sensor location in the visualized model.

This app publishes data to the Marker topic, allowing RViz and other ROS apps to easily subscribe to and visualize this marker topic.

xela_taxel_viz_ahv4 targets the following models:
- uSCuAH (left, right)

Supported models use pre-defined profiles for each model, which can be selected as parameters at startup.

---

---

xela_taxel_viz_ahv4¶

Goals¶

• Visualize /x_taxel_ah tactile data as a 31x26 grid pattern with 368 active cells.
• Support URDF marker mode for Allegro Hand (left/right).
• Reuse the visual style from xela_taxel_viz_2f (circle size/color + arrow direction/length).
• Support real-time updates with best-effort performance.

Inputs¶

• Topic: /x_taxel_ah
• Message: xela_taxel_msgs/XTaxelSensorTArray
• Fields used:
• x_modules[].frame_ids[] (always _joint suffix)
• x_modules[].forces[] (optional)
• x_modules[].taxels[] (primary data)
• x_modules[].temps[] (ignored)

Data Mapping¶

A) Flat index mapping¶

• Use taxel_joint_map_new.yaml to map frame_id -> flat_index (0..367).
• Build a flat array of length 368 per incoming message.
• Matching rule:
• If frame_id exists in the map, place the sample at flat_index.
• If missing, warn and skip.

B) Pattern grid mapping¶

• Pattern grid size: 31 rows x 26 columns.
• Traversal order: top-left, row-major (top->bottom, left->right).
• Pattern YAML contains an index_map matrix:
• -1 for inactive cells
• 0..367 for active cells (flat indices)
• Visualization uses index_map to place each index on the 2D grid.

Value Selection¶

• If forces is present and use_forces_if_present=true, use forces.
• Otherwise, use taxels.
• Baseline is computed per index; visualization uses current - baseline.

Normalization¶

• Use the same normalization model as xela_taxel_viz_2f.
• Ranges (from tech spec):
• X/Y: baseline +/- 1350
• Z: baseline to baseline + 16000
• Z negative deltas are clamped to 0 for normalization.

Visualization Style¶

• Same as xela_taxel_viz_2f:
• Circle radius and color scale with normalized magnitude.
• Arrow length scales with the same magnitude.
• Arrow direction uses XY direction (default).

Parameters (draft)¶

• in_topic, out_topic, frame_id (default x_taxel_ah_viz for grid mode)
• pattern_yaml (pattern file path)
• mapping_yaml (taxel_joint_map_new.yaml path)
• hand_side (left/right prefix rewrite)
• grid_rows, grid_cols, cell_size, origin_x, origin_y
• use_forces_if_present
• baseline_duration_sec
• use_axis_normalization, xy_range, z_range
• circle_*, arrow_*, grid_*, color_*
• marker_stamp_mode, marker_time_offset_sec

Tooling: CSV -> Pattern YAML¶

Recommended workflow¶

1) Export the Excel sheet to CSV (31 rows x 26 cols, empty cells allowed).
2) Convert CSV to YAML using a small Python script.

Example script (Python)¶

python3 - <<'PY'
import csv
import yaml
from pathlib import Path

csv_path = Path('pattern.csv')
out_path = Path('pattern.yaml')

rows = []
with csv_path.open(newline='') as f:
    reader = csv.reader(f)
    for row in reader:
        parsed = []
        for cell in row:
            cell = cell.strip()
            if cell == '' or cell.lower() == 'nan':
                parsed.append(-1)
            else:
                parsed.append(int(float(cell)))
        rows.append(parsed)

payload = {
    'pattern': {
        'rows': len(rows),
        'cols': len(rows[0]) if rows else 0,
        'index_map': rows,
    }
}

with out_path.open('w') as f:
    yaml.safe_dump(payload, f, sort_keys=False)

print('wrote', out_path)
PY

Open Questions¶

• Confirm whether forces should override taxels (default: yes).
• Confirm QoS policy (best-effort suggested).
• Confirm default marker timestamp offset per platform.

Risks¶

• Missing frame_id mapping will leave holes in the grid.
• Performance may degrade if update rate is very high; consider throttling if needed.

---

README¶

RViz2 visualization for /x_taxel_ah (xela_taxel_msgs/XTaxelSensorTArray).
Supports both grid and URDF marker modes for Allegro Hand (AHv4) left/right.

Build¶

cd ~/xela_robotics/02_dev_ws
colcon build --packages-select xela_taxel_viz_ahv4

Run¶

Grid mode (default):

source ~/xela_robotics/02_dev_ws/install/setup.bash
ros2 launch xela_taxel_viz_ahv4 xela_taxel_viz_ahv4.launch.py viz_mode:=grid

URDF mode:

source ~/xela_robotics/02_dev_ws/install/setup.bash
ros2 launch xela_taxel_viz_ahv4 xela_taxel_viz_ahv4.launch.py viz_mode:=urdf

Demo (no /x_taxel_ah source yet)¶

Starts a synthetic publisher and the visualizer.

source ~/xela_robotics/02_dev_ws/install/setup.bash
ros2 launch xela_taxel_viz_ahv4 xela_taxel_viz_ahv4_demo.launch.py

Optional: set demo publish rate

ros2 launch xela_taxel_viz_ahv4 xela_taxel_viz_ahv4_demo.launch.py publish_rate_hz:=10.0

Sim (replayer + server + viz)¶

source ~/xela_robotics/02_dev_ws/install/setup.bash
ros2 launch xela_taxel_viz_ahv4 sim_all_svc_xela_taxel_viz_ahv4.launch.py \
  model_name:=XR23AHLCPP viz_mode:=urdf

RViz2¶

Grid mode fixed frame should be x_taxel_ah_viz.
URDF mode uses the URDF frames (e.g., world/hand_root).

sim_all_svc_xela_taxel_viz_ahv4.launch.py selects RViz configs automatically:

• grid: config/grid_xela_taxel_viz_ah.rviz
• urdf: config/urdf_xela_taxel_viz_ah.rviz

Pattern Conversion (CSV -> YAML)¶

~/xela_robotics/02_dev_ws/install/xela_taxel_viz_ahv4/lib/xela_taxel_viz_ahv4/csv_to_pattern_yaml.py \
  pattern.csv -o pattern.yaml

Key Parameters¶

See config/xela_taxel_viz_ahv4.yaml for defaults.

• mapping_yaml: taxel_joint_map_new.yaml
• pattern_yaml: pattern_lahv4.yaml
• use_forces_if_present: prefer forces when available
• xy_range, z_range: normalization ranges
• hand_side: left|right (prefix rewrite for x_taxel_0_ vs x_taxel_1_)
• marker_stamp_mode: keep|now|zero
• marker_time_offset_sec: default -0.1 for TF stability

Left/Right Notes¶

• model_name:=XR23AHLCPP (left) and model_name:=XR23AHRCPP (right).
• pattern_lahv4.yaml (left) and pattern_rahv4.yaml (right) are selected
  automatically by sim_all_svc_xela_taxel_viz_ahv4.launch.py.

---