---
kind: concept
slug: continuum-robot
title: Continuum robot
summary: >-
  A bendable, snake-like arm used to reach through a narrow access route, such as a natural body opening or a single incision. The arm is driven by wires or by embedded actuation, and its tip is controlled through a kinematic model that is corrected by extra sensing. The trade-off: friction, backlash and stretch make the model inaccurate, and force feedback at the tip is hard to get.
tags: [robotics, continuum-robot, surgical-robotics, medical-robotics]
papers: ["1807.03731"]
related: [magnetic-actuation]
status: draft
updated_at: 2026-11-12
---

## Overview

A continuum robot is a snake-like arm that bends along its length. The review
groups the designs for confined-space surgery into four families: embedded
actuation, linkage-based, wire-actuated and continuum designs [arxiv:1807.03731
§3.1]. The concept covers the last two, and the control problems they share.

The reason for the design is access. Surgery through a natural opening or a
single port means the tools must bend along narrow paths, and many tools share
one access route [arxiv:1807.03731 §2.2]. The surgeon also loses force feedback
and a full view of the scene [arxiv:1807.03731 §2.1]. A rigid instrument cannot
follow such a path, so the arm has to carry the bending itself.

The review treats three problems as the core of the topic. The first is the
mechanical design. The second is the model that maps drive inputs to tip
position, which friction and extension make inaccurate. The third is sensing,
both of the tip position and of the force at the tool. The review does not
propose a design of its own. It maps the field and lists open problems in each
area [arxiv:1807.03731 §4.1].

## Variants

### Wire-actuated continuum arm with closed-loop tip control

Wires run along the arm and pull it into a bend. The review reports that the
control of these arms and of continuum arms uses extra sensing, kinematic
models and calibration [arxiv:1807.03731 §3.2]. The kinematic model alone is
not accurate enough, because friction and backlash in the tendon and backbone
lines, and their stretch under load, change the shape the wires produce. The
review describes compensation methods for these effects [arxiv:1807.03731
§3.3, fig. 6]. Card: [Medical Technologies and Challenges of Robot Assisted
Minimally Invasive Intervention and Diagnostics](/papers/1807.03731).

### Embedded actuation and linkage-based arms

Two other families sit alongside the wire-actuated design. Embedded actuation
puts the drive inside the arm, and linkage-based designs use rigid links
[arxiv:1807.03731 §3.1, fig. 4]. The review describes these families in words
and figure only. It gives no equations or parameters for them, so an
implementer has to go to the cited works [arxiv:1807.03731 §3.1].

### Magnetically held or driven tools

Some surgical tools are held and moved by magnets through the abdominal wall,
without a rigid linkage to the outside [arxiv:1807.03731 §3.5]. The mechanics
and the field-based control are covered on the
[Magnetic actuation](/concepts/magnetic-actuation) page.

## When to use which

The review does not rank these options, so the table reads its section
descriptions against the situation. It does not give measured trade-offs.

| Situation | Use | Why |
|---|---|---|
| Tool must reach through a natural orifice or a single port | Wire-actuated or continuum design | These are the families the review covers for that access route [arxiv:1807.03731 §3.1, §2.2] |
| Tip position must be controlled to a target | Closed-loop control with extra sensing and calibration | The review names this as the control approach for wire-actuated and continuum arms [arxiv:1807.03731 §3.2] |
| Only drive-side readings are available, with no tip sensor | Kinematic model with friction and extension compensation | Friction, backlash and stretch are named as the reasons the model is inaccurate [arxiv:1807.03731 §3.3] |
| Force at the tool-tissue contact is needed | Direct force sensor, or an estimate from motor or actuator signals | The review lists both routes and gives no preference [arxiv:1807.03731 §3.4] |
| Tool must pass through the abdominal wall with no rigid link | Magnetic actuation or fixation | The review covers anchoring, magnetic coupling and electromagnetic actuation [arxiv:1807.03731 §3.5] |

Each row depends on what the review's sources report, and those sources
differ in setup. The review cites numbers from other works and manufacturer
reports, and it runs no experiment of its own [arxiv:1807.03731 §1, §3].

## Implementation sketch

The review does not give a control algorithm for the wire-actuated arm. This
sketch condenses the closed-loop idea that §3.2 and §3.3 describe across
several cited works. It is not a method from the paper. The review gives no
equations for it.

```pseudocode
# Closed-loop tip control of a wire-actuated continuum arm, condensed from
# the review's description [arxiv:1807.03731 §3.2, §3.3]
# Input:  actuator positions (n_a,), desired tip pose
# Output: actuator command (n_a,)

def control_step(actuator_pos, desired_pose, measure_tip):
    # Predict the tip pose from the actuator readings with a kinematic model
    # [arxiv:1807.03731 §3.3]
    predicted_pose = kinematic_model(actuator_pos)
    # Read an external tip measurement when one is available [§3.2]
    measured_pose = measure_tip()
    # Correct the model with the measurement and compensate for transmission
    # losses such as friction and stretch [§3.2, §3.3]
    return compensate(actuator_pos, desired_pose,
                      predicted_pose, measured_pose)
```

The review does not state the form of the compensation, its gains or its
update rate. Those have to be found for each robot.

## Contradictions

None known. The page rests on one review card, and that card reports no
experiments of its own [arxiv:1807.03731 §1].

## Papers

- [Medical Technologies and Challenges of Robot Assisted Minimally Invasive Intervention and Diagnostics](/papers/1807.03731): a review of confined-space surgical robots, covering wire-actuated and continuum arms, their control and sensing, and the open problems in each area.
