ShapeBots: Shape-changing Swarm Robots

INTRODUCTION

• The paper asks how shape-changing interfaces can be distributed and embedded in everyday environments as interfaces become ubiquitous.The authors contrast this prospect with current interfaces that are often large, heavy, and immobile.Source: INTRODUCTION · p. 1
• ShapeBots combines collective and individual shape transformation, extending swarm robots beyond collective rearrangement alone.Figure 2 contrasts the two modes and shows their combination.Source: INTRODUCTION · p. 1, Figure 2 · p. 2
• Individual and collective shape changes support geometric and data displays, tangible control, object actuation, and dynamic physical affordances.Examples include rendering lines or surfaces, clearing a desk, creating a physical constraint, and indicating when a cup is too hot to touch.Source: INTRODUCTION · pp. 1–2
• Each robot uses a 2.5 cm-thin actuator that extends up to 20 cm horizontally and vertically, with modular units enabling varied geometries.The paper demonstrates horizontal and vertical lines, curved lines, 2D area expansion, and 3D volumetric change.Source: INTRODUCTION · p. 2
• The contributions include a concept, a ShapeBots implementation, application scenarios, and a design-space exploration with future research opportunities.The scenarios illustrate potential enhancements to display, interaction, and affordances.Source: INTRODUCTION · p. 2

Swarm User Interfaces

• Swarm UIs use many movable physical elements; allowing individual shape change extends their expressions and interactions beyond collective rearrangement.The paper describes applications including geometric and iconic displays, scientific visualization, physical data representation, accessibility, and tangible controls.Source: Swarm User Interfaces · p. 2

Line-based Shape-Changing Interfaces

• Shape-changing swarm interfaces extend line-based interfaces by combining line deformation with movement and collective shape formation for tangible interactions.Prior work cited includes actuated lines that form objects such as a wristwatch, phone, or ruler.Source: Line-based Shape-Changing Interfaces · p. 2

Modular Shape-Changing Interfaces

• Modular shape-changing systems offer flexible geometries, while limited locomotion and slow reconfiguration constrain some existing approaches.The paper identifies fast locomotion and transformation as key considerations for its self-transformable robots.Source: Modular Shape-Changing Interfaces · p. 2

Definition and Scope

• The paper defines shape-changing swarm robots as collectively movable, self-transformable robots and focuses on their user-interface aspect.It calls this interface aspect shape-changing swarm user interfaces.Source: Definition and Scope · pp. 2–3

2) self-transformation,

• Locomotion brings an interface to users and expands interaction beyond a fixed location, while self-transformation changes an individual element’s shape.The section also places collective behavior among the core aspects of shape-changing swarm interfaces.Source: Figure 3 · p. 3, 2) self-transformation · p. 3

SHAPEBOTS

• ShapeBots demonstrates a swarm of self-transformable robots built around a miniature reel-based linear actuator, robots, tracking, and control.The modular actuator supports individual shape transformation within a collectively movable system.Source: SHAPEBOTS · p. 3

Miniature Reel-based Linear Actuators

• A line-based, modular actuator design supports flexible transformations by combining straight or curved lines and expandable enclosures.The enclosures can support area or volumetric change.Source: Miniature Reel-based Linear Actuators · p. 3

Mechanical Design

• The miniature actuator extends from 2.5 cm to 20 cm, addressing the challenge of fitting substantial deformation capability into a small robot.It uses two motor-driven reels to feed and retract creased polyester sheets; the crease helps resist bending and vertical buckling.Source: Mechanical Design · p. 3
• Actuator extension uses open-loop motor timing and is initialized to zero when a limit switch detects full retraction.The authors report an average target-to-actual length difference below 5 mm; they also note that longer reels risk buckling and sheet friction can cause jams.Source: Mechanical Design · p. 4

Self-Transformable Swarm Robot

• Each robot uses two micro DC motors for forward, backward, and turning motion, with a separate driver and battery-powered control electronics for its linear actuators.The prototype costs approximately 20–25 USD per robot and 6–7 USD per actuator, and includes twelve robots and thirty actuator units.Source: Self-Transformable Swarm Robot · p. 4

Types of Transformation

• The modular actuator units support horizontal and vertical extension, curved lines, volume expansion, and 2D area coverage.The configurations span form, volume, and orientation changes; vertical extension uses double-sided tape to attach the actuator atop the robot.Source: Types of Transformation · p. 4, Figure 6 · p. 4

Electrical Design

• The microcontroller operates separate motor drivers for the wheel motors and linear actuators; synchronously combining signals can operate four actuators but removes independent control.Wi-Fi commands specify wheel PWM values and target actuator lengths.Source: Electrical Design · pp. 4–5

Fiducial Markers and Computer Vision Tracking

• Bottom-mounted ArUco markers let the vision system track robot position and orientation at 60 frames per second.The prototype uses 1.5 cm square 4 × 4 markers, with up to 50 unique patterns.Source: Figure 8 · p. 5

Tracking Setup

• A camera beneath the transparent table tracks a 115 cm × 74 cm area, while a polyester diffuser and lamps improve marker readability.A projector can present graphics without affecting tracking from below.Source: Tracking Setup · p. 5, Figure 9 · p. 5

Control System

• The control system assigns robots to targets using the Munkres algorithm, navigates them with RVO and PID control, then orients and extends them at the target.The web interface lets users draw or upload a target shape and create keyframe animations; the system converts SVG contours into lines for display.Source: Control System · pp. 5–6, Figure 10 · p. 5

Interaction Capability

• Users can place, move, orient, and pick up robots, and the system distinguishes these inputs from unintended events while it is driving them.A single user can manipulate multiple robots with two hands, or multiple users can interact with the robots.Source: Interaction Capability · p. 6, Figure 11 · p. 6

Technical Evaluation

• The technical evaluation measured robot and actuator speed, extension accuracy, pose accuracy, force, latency, and tracking robustness.The authors also compared contour rendering in simulation with non-self-transformable swarm robots at different swarm sizes.Source: Technical Evaluation · p. 6, Method · p. 6, Figure 12 · p. 6

Results

• 151 ms was the total control-loop latency for 10 robots; tracking was lost once every 1.9 seconds on average, corresponding to a 7.9% error rate.The reported latency components were 68 ms for Wi-Fi communication, 39 ms for tracking, and 44 ms for path-planning computation.Source: Results · p. 6
• The moving robot delivered a maximum force of 0.24 N, while vertical actuator load-bearing capability plateaued at about 0.3 N beyond 100 mm extension.An attached actuator delivered the same force as the moving robot up to 130 mm extension; beyond that, the authors observed a significant decrease of approximately 0.9 N.Source: Results · p. 6

Interactive Data Physicalization

• Figure 13 compares ShapeBots and swarm-robot simulations rendering an original SVG image with 30, 40, 50, and 60 robots.The passage describes the comparison setup but does not report which rendering performs better.Source: Figure 13 · p. 7
• Users can move robots to change a physical sine wave’s wavelength, or specify its maximum length to change its amplitude.The paper also describes data representations such as bar graphs, line charts, and star graphs, and a map where robot height represents state population.Source: Interactive Data Physicalization · pp. 6–7, Figure 14 · p. 7, Figure 15 · p. 7

Interactive Physical Display

• Moving one robot in a displayed rectangle makes the others reposition and change length to scale, rotate, or translate the shape.ShapeBots can also render shapes such as a hexagon, fish, or text, and preview a picture-book animal at actual size.Source: Interactive Physical Display · p. 7, Figure 16 · p. 7
• In a CAD preview, changing a box’s height in software updates the robots, while moving robots changes the CAD design.The physical rendering represents the box at actual size.Source: Interactive Physical Display · p. 7, Figure 18 · p. 7

Clean up Robots

• Two robots extend their actuators to wipe debris from a table and clear the user’s workspace.The clean-up scenario illustrates object actuation and use as a physical constraint.Source: Clean up Robots · p. 7, Figure 19 · p. 7

Tangible Game

• In a table-football game, users control robots that act as pinball arms, with their positions and angles synchronized to controller robots.Players use the arms to hit or block the ball toward a goal.Source: Tangible Game · p. 7, Figure 20 · p. 8

In-situ Assistants

• Robots can push a pen to the user, extend to serve as a ruler, or bend to draw a curved line.The ruler’s length is specified by the user.Source: In-situ Assistants · p. 8, Figure 21 · p. 8

Dynamic Fence

• Robots surround a hot coffee cup and change height to form a fence that signals it is not ready to drink; they disperse once it is ready.The paper calls this kind of cue a distributed dynamic physical affordance.Source: Dynamic Fence · p. 8, Figure 22 · p. 8

DISCUSSION AND DESIGN SPACE

• Figure 23 identifies design-space dimensions for shape-changing swarm interfaces and highlights where ShapeBots fits.Figure 24 illustrates future research opportunities.Source: DISCUSSION AND DESIGN SPACE · p. 8, Figure 23 · p. 8, Figure 24 · p. 9

Number of Elements

• More elements can render more expressive and detailed shapes, while greater shape-change capability can help sparse robots fill gaps.Manipulating 10 or 100 elements raises interaction challenges; proposed approaches include grouping, constraints, and higher-level abstraction.Source: Number of Elements · p. 8

Size of Elements

• The paper focuses on 3–4 cm robots, while room-size elements could support dynamic furniture or reconfigurable room layouts.It also considers smaller elements whose aggregated shape changes could alter overall material properties or structure.Source: Size of Elements · pp. 8–9

Input

• The prototype supports place, move, orient, and pick-up inputs; manually extending lines or deforming robots could enable additional applications.Examples include a physical tape measure and tangible demonstration of shape changes.Source: Input · p. 9

Types of Shape Transformation

• The current implementation supports form, volume, and orientation change, while other transformation types could expand its expressiveness and affordances.The paper discusses texture change and combining topologically equivalent and non-equivalent transformations.Source: Types of Shape Transformation · p. 9

Locomotion Capability

• Camera-based tracking confines current locomotion to a tabletop, while future robots could move on walls, ceilings, objects, bodies, or in mid-air.The paper proposes applications including public displays, object moving, and expressive drone displays.Source: Locomotion Capability · p. 9

Materiality of Elements

• The current prototype uses rigid materials; soft swarm robots could add transformations such as changes in volume or stiffness.Fitting actuation into the current robot size remains a technical challenge, and the paper envisions users constructing soft objects that transform.Source: Materiality of Elements · p. 9

Connectability

• Connecting swarm elements could enable graspable 3D shapes that dynamically change form or scale, but the current prototype elements do not physically connect.Such shapes could provide haptic proxies for virtual objects or previews of 3D designs.Source: Connectability · p. 9

Actuation Method

• Reducing robots below 1 cm remains technically challenging because each prototype contains locomotion, actuation, and battery components.External actuation is discussed as a possible way to address some of these problems.Source: Actuation Method · p. 9

Representation

• Physical representations can be combined with graphics, including projection-mapped information, LED color, and mixed-reality overlays on physical CAD previews.Haptic feedback for virtual or augmented reality is proposed as a research direction.Source: Representation · pp. 9–10

Strength of Actuation

• The actuator moves only lightweight objects and is not stable enough for heavy-object actuation or resistance; stronger actuators could support larger or load-bearing applications.The discussion mentions adjustable bookstands, footrests, and room-scale robots as possibilities.Source: Strength of Actuation · p. 10

Authoring of Behaviors

• The GUI authoring tool supports keyframes or scripts, but animation and interactive authoring remain primitive, while programming by demonstration poses abstraction challenges for large swarms.The authors identify improvisational authoring in physical space as an important direction.Source: Authoring of Behaviors · p. 10

LIMITATIONS AND FUTURE WORK

• Scaling from the twelve-robot prototype to hundreds raises tracking-latency and collision concerns; retracting reels before movement can disrupt animation.The cited detection time is 1.4 ms in 1080p; the authors note it can maintain the current frame rate with 10–20 robots, while latency may cause problems above 100.Source: LIMITATIONS AND FUTURE WORK · p. 10
• The prototype’s force output is relatively weak for haptics or heavy-object actuation, and the authors suggest friction, reel materials, or scissor mechanisms as possible improvements.Source: LIMITATIONS AND FUTURE WORK · p. 10
• The prototype emphasizes output over rich touch or manual-reel input; more expressive geometries are constrained by size, resolution, and two-actuator control, while batteries require hourly recharging.The authors also identify larger-area tracking outside a specially designed table as future work.Source: LIMITATIONS AND FUTURE WORK · p. 10

• ShapeBots demonstrates that self-transformable robots can combine individual shape change with collective behavior to support varied shape representations and interactions.Source: CONCLUSION · p. 10