Hierarchical Admittance-Impedance Control for Mobile Manipulator Load Transfer

Minghao Xu, Thushara Sandakalum, Christina Dao Wen Lee, Marcelo H. AngarXiv 2026 · arXiv

TL;DR

This paper introduces a three-layer hierarchical control framework that allows a mobile manipulator to act as a collaborative partner for heavy lifting. By assigning distinct "interaction roles"—using the arm to provide stable load support (impedance) while a supervisor translates human touch into motion (admittance)—the system enables a person to guide a heavy suitcase across a room with just a fraction of the effort normally required.

What Problem Does This Paper Address?

Physical human-robot collaboration (pHRC) is a "holy grail" for industries like airport baggage handling, where workers face high physical strain from repetitive lifting. Ideally, a robot should shoulder the weight of a payload while remaining completely responsive to a human’s intuitive pushes and pulls.

However, achieving this on a mobile manipulator is technically difficult. You need the robot arm to be "stiff" enough to hold a heavy object without dropping it, yet "soft" enough to move when a human nudges it. Furthermore, the mobile base must coordinate its movement with the arm so the robot doesn't overextend itself or become unstable. Most existing solutions either focus on purely autonomous trajectories (which lack human intuition) or require complex, unified whole-body torque controllers that are difficult to implement on hardware with different control interfaces.

Methods

The authors propose a hierarchical architecture that separates the control problem into three distinct layers, as illustrated in the control architecture diagram:

![Fig.

  1. Hierarchical admittance-impedance control architecture for human-guided load transfer with workspace-aware mobile-base assistance.](/figures/art_ed03c17fe9cf4affa5462907b3dd7b0f)
  2. Inner Cartesian Impedance Controller: This layer runs on the Franka Research 3 (FR3) arm. It treats the robot-object interface as an impedance. It regulates the relationship between the arm's displacement and the restoring force, essentially acting like a programmable spring-damper system that keeps the load stable around a specific point.
  3. Outer 6D Admittance Supervisor: This layer handles the human-robot interface. It treats the system as an admittance. It measures the force (wrench) applied by the human and calculates how the robot should move in response. This "motion intent" is then passed down to the inner controller as a moving equilibrium point.
  4. Workspace-Aware Mobile Base Assistance: This layer coordinates the wheels. It monitors the arm's extension; if the human guides the load beyond a certain reach, the base automatically drives forward to "assist." It also includes a "frontal load alignment" feature that rotates the base to keep the payload directly in front of the robot, ensuring better stability and a more intuitive experience for the user.

Datasets

The paper does not use standard machine learning datasets. Instead, the system was validated through a custom suitcase-transfer experiment. The setup involved a 3.3 kg suitcase (generating ~32.4 N of gravitational force) which a human operator guided from a starting position to a target table using the mobile manipulator.

Key Insights and Ideas

  • Role Assignment is Key: The "aha moment" here is the interaction-role interpretation. By deciding that the robot-object interface should be an impedance and the human-robot interface should be an admittance, the authors created a stable way to "buffer" human input.
  • Handling Heterogeneous Hardware: Most mobile manipulators have a "split personality": the arm often accepts torque commands, while the base only accepts velocity commands. This hierarchy bridges that gap without needing a complex, unified mathematical model of the entire robot.
  • Frontal Alignment for Safety: Keeping the load in the "frontal support region" (as shown in Figure 1) isn't just about ergonomics; it reduces lateral side-loading on the arm's joints and keeps the center of gravity within a safe footprint for the mobile base.

![Fig.

  1. Physical roles during human-guided load transfer using a mobile manipulator.](/figures/art_1e3e36e8547c48a6be5e632c2d8f0e10)

Key Results

The experimental results demonstrated a significant reduction in human effort:

  • Effort Reduction: While the robot supported a stable load of approximately 35.40 N, the human operator only needed to apply an average guidance force of 7.43 N (roughly 21% of the load weight) to move it.
  • Workspace Extension: The suitcase was transported over a total distance of 2.44 meters, far exceeding the reach of a stationary arm.
  • Configuration Stability: Thanks to the base assistance, the arm's extension remained remarkably consistent, staying within a narrow range of 0.57m to 0.645m throughout the entire transfer.
  • Intuitive Alignment: The absolute load angle stayed within an average of 10.31°, confirming the robot successfully kept the suitcase in front of itself during the walk.

Limitations and Future Work

While effective, the current system relies on manual parameter tuning for the admittance and impedance gains. The authors note that performance is highly sensitive to these selections and suggest that future work should involve optimization-based or learning-based methods to tune these parameters automatically based on the task or the specific human user.

Additionally, the current "grasping" phase is still somewhat manual. The authors aim to integrate autonomous perception and grasp planning so the robot can identify and pick up a suitcase on its own before handing over the "steering" to the human worker.

Summary updated

Abstract

This paper presents a hierarchical admittance-impedance controller that combines an inner Cartesian impedance controller, an outer 6D admittance supervisor, and workspace-aware mobile-base assistance for guided load transfer. The method is motivated by assistive tasks such as airport baggage handling, where a robot should support a payload while a human worker guides its motion through small physical forces. Following an interaction-role interpretation of admittance and impedance, the torque-controlled Franka Research 3 manipulator behaves as a Cartesian impedance at the robot-object interface, providing compliant load support around a commanded equilibrium pose. At the human-facing interaction level, the outer admittance supervisor converts measured human wrench input into smooth Cartesian reference motion. The workspace-aware assistance layer shares the admittance-generated translational motion between the manipulator and a velocity-controlled base when larger motion is required, while front-side load alignment keeps the carried object near the robot's frontal region for safer and more intuitive guidance. The system was implemented in ROS and validated through a suitcase transfer experiment, showing that the robot can support the load while allowing the human to guide the transfer with relatively little effort.
BibTeX
@misc{xu2026hierarchical,
  author = {Minghao Xu and Thushara Sandakalum and Christina Dao Wen Lee and Marcelo H. Ang},
  title = {{Hierarchical Admittance-Impedance Control for Mobile Manipulator Load Transfer}},
  year = {2026},
  eprint = {2611.24771},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.24771},
}
Figure