Implementation:Google deepmind Dm control Walker Info

Knowledge Sources	Google_deepmind_Dm_control
Domains	Robotics, Motion Capture
Last Updated	2026-02-15 04:00 GMT

Overview

The WalkerInfo class encapsulates routines for modifying a walker's MJCF model to match the proportions and marker layout of a motion capture subject.

Description

WalkerInfo wraps a mocap_pb2.Walker protobuf and provides methods to adapt a walker model so that it physically matches the motion capture data. The class bridges the morphological differences between a generic walker model and a specific motion capture subject.

The rescale_walker method adjusts body subtree proportions by iterating over the protobuf's SubtreeScaling data. For each subtree, it computes a position factor either from parent_length (dividing by the current body position norm) or directly from size_factor, then delegates to rescale.rescale_subtree to recursively scale positions and sizes. If the protobuf specifies a target mass, the method compiles a temporary physics instance to read the current subtree mass, computes a mass factor, and scales all body inertial masses, geom masses, or geom densities accordingly.

The add_marker_sites method creates small sphere sites on walker bodies at positions specified by the protobuf's marker data. These sites are used for motion capture marker tracking and are styled using a dedicated mocap default class for consistent visualization. A compatibility check method ensures the walker's model type matches the protobuf's model enum before any modifications.

Usage

Use WalkerInfo when you need to adapt a walker model to match specific motion capture subject proportions. It is typically invoked through Trajectory.configure_walkers() rather than used directly. This is essential for accurate motion tracking where the simulated walker must match the physical dimensions of the recorded subject.

Code Reference

Source Location

• Repository: Google_deepmind_Dm_control
• File: dm_control/locomotion/mocap/walkers.py
• Lines: 1-97

Signature

class WalkerInfo:
    def __init__(self, proto):
    def check_walker_is_compatible(self, walker):
    def rescale_walker(self, walker):
    def add_marker_sites(self, walker, size=0.01, rgba=(0., 0., 1., .3),
                         default_to_random_position=True, random_state=None):

Import

from dm_control.locomotion.mocap.walkers import WalkerInfo

I/O Contract

Inputs (WalkerInfo.__init__)

Name	Type	Required	Description
proto	mocap_pb2.Walker	Yes	Protobuf message containing walker model type, scaling data, mass, and marker info

Inputs (add_marker_sites)

Name	Type	Required	Description
walker	Walker	Yes	The walker entity to add marker sites to
size	float	No	Radius of marker site spheres (default: 0.01)
rgba	tuple	No	Color of marker sites (default: blue, semi-transparent)
default_to_random_position	bool	No	If True, use random position when marker has no position data (default: True)
random_state	numpy.random.RandomState	No	Random state for generating default marker positions

Outputs

Name	Type	Description
rescale_walker()	None	Modifies the walker's MJCF model in-place to match mocap proportions
add_marker_sites()	list of mjcf sites	List of created site elements; also sets walker.list_of_site_names

Usage Examples

from dm_control.locomotion.mocap.walkers import WalkerInfo

# WalkerInfo is typically used through Trajectory.configure_walkers()
trajectory = loader.get_trajectory('clip_001')
trajectory.configure_walkers(walker)

# Direct usage (less common):
walker_info = WalkerInfo(walker_proto)
walker_info.check_walker_is_compatible(walker)
walker_info.rescale_walker(walker)
marker_sites = walker_info.add_marker_sites(walker)