Elastic robotic structures: a multidisciplinary framework for the design and control of shape-morphing elastic system for architectural and design applications

1.1 Background

Architects, designers and engineers have for a long time imagined intelligent, adaptable and interactive environments (Negroponte 1976; Kolarevic 2009; Fox 2016), exploring the relationship between human beings and their surroundings (Farahi and Leach 2023).

While technical challenges previously restricted these ideas to limited implementations in terms of scale and behaviors, recent advances in architecture, engineering, and robotics have enabled the reconsideration of these visions and the development of new shape-morphing structures. These interactive systems can enhance everyday activities (Salvendy 2012), creating new types of experiences where objects and environments change their shape and appearance to accommodate different functions, communicate information (Coelho and Zigelbaum 2011), and respond to various parameters, such as human, environmental, functional, and structural stimuli. Recognition of the impact that interactive environments can have on our lives has led to an increasing cross-disciplinary focus on shape-morphing technologies, with applications spanning wearables (Farahi 2016), objects (Togler et al. 2009), robotics (Bertoldi et al. 2017), assistive devices (Choi et al. 2021), furniture (Mhatre et al. 2021; Sareen et al. 2017), and building systems (Lienhard et al. 2011).

1.2 Research problem

When designing shape-morphing systems, the selection of materials and mechanisms to achieve the desired deformation is crucial. Conventional kinetic systems rely on rigid body kinematics, achieving motion through the connection and actuation of multiple rigid parts (Schleicher et al. 2011). A different design strategy is to leverage highly deformable elastic materials to achieve motion. ‘Material interactivities’, as discussed by Tibbits, Farahi and Leach (2023), describe a design approach that harnesses physical material behaviors to develop intelligent systems capable of autonomous assembly and shape changes. While there are ongoing advancements in the field of interactive architectural systems, particularly in the development of new human–environment interactions (Ghandi et al. 2021), smart buildings (Al Dakheel et al. 2020) and living structures (Beesley 2020), one of the primary challenges remains realization of systems that can achieve a wide range of shapes, multiple states of equilibrium, while also minimizing energy and material use. One of the key difficulties is implementing large-shape transformations in continuously operating structures without compromising stability.

Elastic systems address this issue by enabling substantial shape changes with minimal material and actuation requirements (Schleicher et al. 2011). Despite their adaptability and ability to deform reversibly under different loads without permanent damage (Lienhard and Knippers 2015), the highly non-linear behavior of elastic materials makes them challenging to predict and control at large scales. Elastic thin-walled systems, such as bending-active and tensile pneumatic structures in architecture and engineering, create lightweight, structurally efficient structures where geometry arises from the equilibrium of forces and material behavior ((Lienhard 2014; Knippers et al. 2011)). Advances in computational design tools have broadened the scope for designing these systems (Suzuki 2020), yet they typically remain static once constructed. Only a limited number of these systems have shown kinetic capabilities, hindered by the challenges of managing large deformations and the absence of control methods in traditionally static architecture (Soana et al. 2020).

The exploration of elastic shape-morphing systems spans fields such as soft robotics, human–computer interaction (HCI), and structural mechanics. Traditionally, these disciplines have focused on small-scale, specific applications. However, there is a growing effort to expand creative boundaries (Benli 2023) and scale up to human or room sizes (Li et al. 2022; Siéfert et al. 2020).

Despite these efforts, design possibilities, shape-changing capabilities, and behaviors remain limited due to the lack of integrated methodologies that facilitate broad design explorations, stability and control.

1.3 Contribution

Further explorations in this area are essential, as current discipline-specific tools and systems are limited compared to the vast potential of soft shape-morphing components in the field of interactive systems at human and architectural scales.

Given the conceptual and technical complexity of designing interactive objects and environments that engage with humans, there is a clear need for multidisciplinary approaches that integrate methods, perspectives, and expertise from the realms of design, architecture, engineering, and robotics. This paper introduces the concept of Elastic Robotic Structures (ERS) and a multidisciplinary framework for their design, realization, characterization and control. Focusing on developing lightweight, shape-morphing systems that are robotically actuated, the ERS framework and its implementation through realized projects (Fig. 1) represent a collaborative effort by multidisciplinary teams of academics, researchers and students from architecture, engineering and robotics, led by the authors. ERS framework is structured through specific steps and methods, integrating the flexibility and interactivity of design methodologies with the reliability of robotic solutions (Fig. 2).

Images showing a selection of ERS developed with the multidisciplinary framework presented. (1) Self-Choreographing Network, 2018, University of Stuttgart; (2) ELAbot, 2020, Department of Computer Science (CS), University College London (UCL); (3) ELAbody, 2021, Venice Architecture Biennale, CS, UCL. The following projects have been developed within the Architectural Design Program (AD), BPro, at the Bartlett School of Architecture (BSA), UCL: (4) LOOPS, 2020/21; (5) Elastic Choreographies, 2020/21; (6) WINGS, 2021/22; (8) TPop, 2021/22; (9) EMObot, 2022/23; (10) A dance of other bodies (ADOOB), 2022/23. The following projects have been developed within the Architectural Computation Program (AC), BPro, BSA, UCL: (7) Programmable Pneumatic system (PPS), 2021/22; (11) Aerobot, 2022/23; (12) Tensformer 2022/23. Full Authors and Team list in the acknowledgments

Full size image

ERS multidisciplinary approach integrates methods used to design and fabricate elastic lightweight structures in architecture and structural design with soft robotics and interactive systems. The figure includes images of: (1) LOOPS tensile system, (2) the Multihalle Mannheim, Frei Otto (Möller and Fischer 2018), (3) a robotic system from Elastic Choreographies (Sabery et al. 2023), (4) a soft robot, (5) ADOOB human-interaction strategy. The bottom image shows a diagram of the cyber-physical control and behavior of ELAbot (Soana et al. 2020)

Full size image

Hence, it leverages the strengths of tools from various disciplines. The main focus has been on identifying how different disciplines approach similar problems, selecting methods and establishing communication between these different environments and procedures.

The presented framework is developed to be modular and flexible, facilitating the integration of unique functionalities specific to individual ERS projects developed over. The adaptivity of the developed ERS is achieved by strategically integrating elastic materials with actuators, enabling multiple equilibrium states through controlled elastic deformations. ERS interact with humans and adapt to diverse conditions, including design, structural and environmental parameters.

ERS systems are examined and categorized based on different parameters such as design scopes, structural-actuation principles, and behaviors.

They include bending-active, tensile, pneumatic and hybrid structures, actuated by various mechanisms such as motorized joints, variable tendons and air sources. Given the challenges of controlling ERS, operations are embedded within a cyber-physical network that integrates methods from different disciplines to continuously exchange and process simulation and physical data to achieve desired behaviors. ERS behaviors, therefore, emerge through dynamic negotiations of interoceptive and exteroceptive parameters. ERS’s versatility extends to applications in robotics, wearables, interactive objects and adaptive building systems (Fig. 3).

Applications of developed ERS include: (1) adaptive buildings and structures  (Elastic Choreographies), (2) interactive furniture (WINGS), (3) wearable assistive devices such as haptic mask for visually impaired people (CyberVision), (4) crawling robot (LOOPS), and (5) robotic actuators (TPop)

Full size image

1.4 Paper structure

Section 2 provides a critical literature review, presenting the conceptual and technical context of ERS in different disciplines, highlighting the challenges and opportunities offered by employing a cross-disciplinary approach. Section 3 proposes the fundamentals required to understand the development of the multidisciplinary framework introduced for ERS development. It outlines the overarching ERS design approach before presenting the multidisciplinary framework for their design and control, and how it has been implemented in the realization of several ERS. Conclusions are drawn in Section 4.

2.1 Lightweight structures in architecture and engineering

Elastic systems in architecture, such as bending-active, tensile, pneumatic, and hybrid structures, generate efficient, lightweight structures. Bending-active systems, forming curved geometries from elastic deformation of straight or planar elements, can revert to their original shape within elastic limits, making them suitable for compliant mechanisms (Knippers et al. 2011; Lienhard 2014; Suzuki 2020). Tensile systems, also defined as form-active (Engel 1999), include cables, pre-stressed and air-supported membranes (Lienhard and Knippers 2015). They rely on tensile forces for structural integrity, allowing large spans and complex shapes with minimal material use, resulting in lightweight designs. Hybrid systems integrate these elements, with global shapes emerging from force equilibrium between structural elements.

These systems evolved from vernacular constructions to cost-effective solutions to advanced structures, influenced by pioneers such as Frei Otto and Buckminster Fuller (Lienhard and Knippers 2015; Oliver 2007; Leonhardt 1940; Chi and Pauletti 2005; Pauletti 2010). More recent applications range from temporary structures to innovative building envelopes. Significant advancements have been made in developing new systems and computational tools, driven by academic research (Lienhard and Gengnagel 2018), such as the work of Lienhard, Suzuki, and Ahlquist at the University of Stuttgart (Lienhard et al. 2013; Suzuki 2020; Ahlquist et al. 2013, 2015; Ahlquist 2015, 2016), and projects like the Tower, CITA (Thomsen et al. 2015).

The shape of elastic structures depends on the equilibrium of forces and material behavior (Stimpfle 2008; Stranghöner et al. 2016; Stranghöner et al. 2023). Consequently, their design cannot rely on conventional geometrical approaches but must emerge through form-finding processes (Pauletti 2010), where the final shape depends on the interplay between material behavior, geometrical and topological constraints, and force distribution. As a result, physics-based design tools become fundamental for designing elastic structures. An appropriate selection of tools is particularly relevant in design disciplines where, unlike purely engineering problems, the exploration of form is a foundational aspect. This underscores the substantial difference between processes employed in architecture and design, and those used in more narrowly focused engineering problems (Mueller 2017). This distinction arises from the fact that specific engineering problems can often be assessed through quantitative criteria, leading to the preferences for simple shapes, whereas design products necessitate a multi-criteria and qualitative evaluation, often requiring the development of complex shapes. In a multidisciplinary context, this problem is more intricate, requiring the assessment of different tools and design approaches.

Form-finding for elastic structures mainly uses numerical techniques, as other methods such as empirical, experimental, and analytical (Huerta 2006; Boller and Schwartz 2020) are less reliable for complex, large-scale designs (Suzuki 2020). These numerical methods, including finite-element modeling (FEM) and physics-based tools, model forces and material behavior to define the final shape. While FEM is a structural analysis standard (Quinn 2020), it struggles with interactivity (Brandt-Olsen 2016a), which is important for shape-changing systems.

Physics-based modeling tools, originating from computer graphics, offer fast simulations but lack accuracy in real material behavior, often requiring FEM validation (Suzuki 2020). To bridge this gap, architects and engineers developed tools for interactive design with more realistic material representation. An example is Kangaroo Physics, a Rhino 3D and Grasshopper plug-in by Daniel Piker (2017). It uses dynamic relaxation (DR), a lightweight structure analysis technique from the 1960s (Barnes 1999), with goal-based projections and energy minimization via ADMM (Quinn 2020). Kangaroo, an open-source tool, was enhanced for specific needs like the TOWER project at CITA (Deleuran et al. 2015), and Cecilie Brandt-Olsen’s K2 Engineering extension incorporates real stiffness values for precise structural analysis (Brandt-Olsen 2016b).

Current research in computational design is addressing simulation challenges to develop tools for the inverse design of pneumatic shapes (Panetta et al. 2021) and deployable grid shells (Becker et al. 2023). Although there is growing interest in creating computational tools that can accurately simulate large deformations in elastic systems interactively, most of these approaches have not yet considered the prediction and control challenges of a continuously changing physical system.

2.2 Architectural soft robots and human-scale interactive systems

This section provides a historical overview of how interactive and adaptive large-scale systems have evolved. Following the conceptual introduction, the focus shifts to soft adaptive projects at the architectural scale.

The development of interactive architectures and intelligent environments begins with the influence of Cybernetics in the 1960s. In the 1970s, Nicholas Negroponte introduced the concept of ‘soft architecture machines’ (Negroponte 1976), envisioning architectural spaces that adapt to users through computational processes. In his book, he also explored the potential of soft material systems, exemplified by Sean Wellesley–Miller’s work at MIT, to develop adaptive structures capable of self-forming and ‘even walking’ if designed appropriately. Architects such as Cedric Price and John Frazer, influenced by Gordon Pask, further explored these concepts, leading to innovative interactive designs (Steenson 2022). The architecture machine group (AMG) at MIT epitomized this movement, offering a new interdisciplinary perspective on how artificial intelligence (AI) could interface with the physical environment.

The growing interest in lightweight material systems, influenced by Frei Otto’s work, led to innovative structural systems and visionary projects challenging static architecture norms (Otto et al. 1991; Möller and Fischer 2018). Coop Himmelb(l)au’s soft spaces aimed at creating immersive environments, where sensory elements and tactile interactions fostered meaningful connections between the built environment and humans. Visionary projects such as David Greene’s Living Pod and Michael Webb’s cushicle introduced innovative models of mobile soft architecture. The exploration of architectural soft machines persisted through projects such as Mark Fisher’s Automat and Dynamite, utilizing pneumatic cells for dynamic transformations (Wihart 2015). While most projects remained theoretical or on a prototypical level, late 1990s technological advancements in ubiquitous computing led to practical revisions (Yiannoudes 2016). Kas Oosterhuis’ Hyperbody Research Group at Delft University focused on developing soft-actuated structures for real-time shape changes in response to internal and external inputs (Oosterhuis 2005). This “emotive” architecture aimed to adapt functions and enhance human experience, leading to the “Muscles projects” (Biloria 2010; Oosterhuis and Biloria 2008).

Since the 1960s, interactive architecture has seen a proliferation of projects, from kinetic structures to computationally enhanced installations, with notable contributions by Beesley (2020), Farahi and Leach (2023) and Sabin et al. (2020). Concurrently, engineers have developed “Intelligent Environments” for domestic settings (Yiannoudes 2016), and architectural robots have become integral to the built environment through interdisciplinary collaboration (Gross and Green 2012; Green 2016; Bier 2011). Henriette Bier’s work has extended this concept, emphasizing a life-cycle approach to robotic buildings (Bier 2015; Bier and Mostafavi 2016). Research in elastic adaptive systems has also advanced, exemplified by projects such as flectofin and flectofold at ITKE University of Stuttgart (Lienhard 2014; Lienhard et al. 2011; Schleicher et al. 2011; Körner et al. 2017), and the Ocean Pavilion in South Korea. Axel Killian’s exploration of autonomous architectural robots, such as the Flexing Room Architectural Robot, has contributed to this field (Killian 2017). The Self-Assembly Lab at MIT’s work on large-scale transformable structures using flexible biaxial braided geometries marks another significant development (Sparrman et al. 2017).

The growing architectural interest in soft-actuated systems, driven by their design qualities of softness, has led to numerous research projects. These projects utilize soft robotics methods to envision interactive architectural elements, often at a modular or small scale (Decker 2015; Wihart 2015; Kapelonis 2018).

Recent advancements in elastic kinetic systems at a human scale have focused on developing new fabrication and characterization methods. The ITECH Research Demonstrator 2018–19 introduced large-scale compliant folding mechanisms, while ‘Curated Deformations’ demonstrated shape-changing surfaces through the elastic deformation of bending-active pneumatic components (Körner et al. 2021; Mühlich et al. 2020). In addition, research in architectural engineering has concentrated on the mechanical characterization of hybrid tensile and bending systems in multiple states (Puystiens 2015). Physicists and engineers are also exploring large-scale multi-stable soft systems, with a focus on buckling-induced deployable structures (Mhatre et al. 2021), pneumatic actuation in origami structures (Melancon et al. 2021), and planar patterned fabrics (Siéfert et al. 2020). At MIT’s Medialab, multidisciplinary teams are working on projects like Printflatables pneumatic structures, integrating soft robotic principles into furniture design (Sareen et al. 2017). The scaling of soft robotics to human sizes (Usevitch et al. 2020; Melancon et al. 2021)—and the development of shape-changing interfaces and large-scale projects such as the Commotion Bench and lifTiles—are effectively merging architectural and design concepts with robotic techniques (Grönvall et al. 2014; Suzuki et al. 2020).

Given the challenges in controlling elastic kinetics, research has largely focused on isolated aspects such as design, material-actuation studies, fabrication, or robotic control. Consequently, the geometric freedom of these systems remains limited, lacking unified and appropriate methods for designing and controlling large-scale deformations.

2.3 Soft robotics and shape-changing interfaces

Soft robotics is the field that explores the design and control of robots made from compliant materials. Soft robots have several advantages over traditional rigid robots, given their inherent compliance and the ability to achieve complex motions from simple inputs (Rus and Tolley 2015). These robots are safe for human interaction due to their use of soft materials (Polygerinos et al. 2017) and can be actuated pneumatically or through variable-length tendons (Rus and Tolley 2015; El-Atab et al. 2020; Wang and Chortos 2022; Li et al. 2018). These robotic systems are designed for a wide range of applications, including locomotion (Sun et al. 2021), manipulation (Shintake et al. 2018) and medical (Hsiao et al. 2019). Due to the non-linearity and time dependency of compliant material behaviors, controlling such robots can be challenging. Control strategies for soft robots include open-loop, closed-loop and autonomous control. Wang and Chortos (2022) describe these as: (1) open-loop control, where control data are calculated from a robotic model and input value, then sent to the actuator(s), with complexity increasing with more actuators and behaviors; (2) closed-loop control, which uses feedback from sensors in its first level, and global configurations including kinematic and dynamic models in its second level; (3) autonomous control, which employs optimization and machine learning to automatically generate control data.

While widely used in soft robotics, the complexity of the system, its shape, and its desired behaviors influence the level of control implementation required (Wang and Chortos 2022).

Methods in soft robotics, designed and tested primarily for small-scale, specific applications, are limited and require significant adaptation for use outside traditional robotics. As interest grows in applying soft robotics creatively to everyday interactive devices, a multidisciplinary approach becomes essential. Intersectional efforts between fields such as HCI, materials science, design and robotics led to a series of projects identified as shape-changing interfaces (Coelho and Zigelbaum 2011), also known as actuated and organic user interfaces (Parkes et al. 2008; Poupyrev et al. 2007), shape-transformable materials (Lee et al. 2022), flexible mechanical metamaterials (Bertoldi et al. 2017) and computational composites (Vallgårda and Redström 2007).

Shape-changing interfaces, which use physical shape modifications for input or output, are applicable to various objects, enabling dynamic user interactions (Follmer et al. 2013). These interfaces cover a broad spectrum of technologies and applications like deformable displays and adaptive materials, enhancing user interaction through physical changes and object manipulation (Alexander et al. 2018). Made from diverse materials and mechanisms, they allow for changes in topology, texture and permeability (Coelho and Zigelbaum 2011). The design of these interfaces involves considering the dynamics of shape change, interaction and system purpose, categorizing changes in orientation, form, volume, texture, viscosity, spatiality, adding/subtracting and permeability (Rasmussen et al. 2012).

Evaluating these systems poses challenges, leading to the development of comparison metrics (Roudaut et al. 2013), prototyping tools (Kwak et al. 2014) and transformation analysis (Rasmussen et al. 2012). Further, attempts have been made to identify expressive parameters that describe the purpose of different changes and how they are perceived by the user: from hedonic or non-instrumental to communication and feedback. The “Grand Challenges in Shape-Changing Interface Research” aimed to coherently analyze the current state of the art and new directions of the field (Alexander et al. 2018). They identified the main motivation behind shape-changing interfaces, such as the ability to: (1) adapt to tasks, users and environment; (2) communicate information; (3) provide hedonic and symbolic purposes; (4) augment users; and (5) simulate objects (Alexander et al. 2018). In the same year as Sturdee, Alexander proposed another classification and analysis of shape-changing interfaces (Sturdee and Alexander 2018). They proposed “a meta-analysis of shape-changing design theory, a detailed database of shape-changing prototypes, and a categorization of types of shape-changing interface (Enhanced 2D, Bendable, Paper and Cloth, Elastic and Inflatable, Actuated, Liquid, Malleable, and Hybrid)”. They aimed to make a repository of information on shape change so that designers could refer to it to identify the essential characteristics and constraints of any existing shape-changing prototypes. Given the multidisciplinary space, Qamar et al. have discussed the evolving relationship between HCI and material science, emphasizing the critical synergy between the two fields in designing shape-changing devices (Qamar et al. 2018).

In the realm of shape-changing interfaces and systems, there is a wide array of materials and mechanisms employed. For the purpose of this research, the emphasis is placed on systems that undergo elastic deformation, such as stretchable, bending and inflatable configurations. In the last two decades, there has been a growing focus on enhancing everyday activities through the integration of interactive soft objects. Psychological research has indicated that the effective management of mental workload can be facilitated by incorporating various types of input (Salvendy 2012). Haptic feedback has proven to be a viable alternative to visual and auditory feedback (Scott and Gray 2008). This led to different studies and designs of haptic objects, including car parts (Gaffary and Lécuyer 2018; Telpaz et al. 2015; Enriquez et al. 2001; Birrell et al. 2010; Audi 2017). This novel approach aims to provide feedback to drivers more effectively, and create novel user experiences. The development of robotic shape-changing furniture can have a range of applications, from therapeutic (Yaqun et al. 2009) to stabilizing postures (Li and Fujimoto 2019), to enhanced design features.

Other examples of interactive soft objects include an inflatable mouse (Kim et al. 2008), an adaptive voice recorder (Zigelbaum et al. 2008), adaptive buttons (Harrison and Hudson 2009), and a shape-changing water dispenser designed with a lifelike appearance (Togler et al. 2009). The dispenser serves as an effective tool for encouraging users to reflect more thoughtfully on their daily water consumption.

Many of these devices aim to communicate information in an embodied manner. Aspire is a clippable, mobile pneumatic-haptic device for breathing rate regulation (Choi et al. 2021). Respire (Jain et al. 2023) is an interactive biofeedback art installation that promotes self-awareness and interpersonal connectedness by encouraging playful interactions with its shape-shifting elements.

While the cited projects have each been developed for a specific purpose, one of the main focuses in this field has been on the development of enabling technologies that could be used for different purposes (Coelho et al. 2008; Parkes and Ishii 2010; Wang et al. 2018; Tahouni et al. 2020; Stanley and Okamura 2016; Iwata et al. 2005; Ou et al. 2016; Han et al. 2020). Despite ongoing efforts, the field of shape-changing interfaces is still quite fragmented; there are no unified guidelines on design methodologies that enable flexibility and freedom in the conceptual phase, while providing integrated control solutions for complex shape changes and responsive behaviors at different scales.

This section is organized into three sub-sections. The first, ‘ERS Design Approach,’ introduces the fundamental concepts and strategies essential for understanding ERS. The second, ‘ERS Multidisciplinary Design and Control Framework,’ presents the detailed framework developed for the design and control of ERS, building upon the previously discussed ERS concepts. The final sub-section, ‘Overview of Built ERS,’ offers a chronological review of selected ERS projects, categorizing them based on various parameters to illustrate the application of ERS concepts in different contexts.

3.1 ERS design approach

Designing shape-morphing ERS at human and architectural scales requires a fundamental shift away from traditional design approaches. Typically, built systems are designed to avoid large deformations, in processes that clearly separate the digital conceptualization phase from the physical materialization.

This approach often leads to the creation of oversized, rigid elements, as it does not encourage active engagement with material behaviors in the design process. In the case of lightweight elastic structures, the final form emerges from material behaviors, determining not just the aesthetic appearance of the system but also its structural state. This approach differs from traditional methods, marking a significant departure from the standard practice of focusing on rigidity and permanence while avoiding active material behaviors. The design of continuously adaptive systems like ERS requires new methods that enable the generation of multiple states, which are significantly distinct from each other, based on structural and material properties. In addition, it necessitates the development of decision-making and control strategies for post-development behavior. Therefore, there are two main aspects to the ERS design process: selection of the structural response and actuation strategy, and selection of the target system behaviors (Fig. 4).

1. 1.

   Structural response and actuation strategy: ERS systems consist of the strategic integration of actuation strategies in elastic structural systems. ERS structural responses comprise: bending-active, tensile systems, pneumatic and hybrids (Fig. 4, top). Actuation methods include topological changes, variable-length cables, incremental tensioning mechanisms and pneumatics (Fig. 4, middle). Topological changes refer to alterations in the position or angle of a structural point, such as connections between elastic beams in a bending-active grid-shell or network, and anchor points of bending or tensile elements. Variable-length cables and incremental tensioning involve adjusting the length of cables connected to either bending elements or tensile surfaces. Pneumatic actuation uses air pressure to shape tensile materials into specific forms. The selection of a structural response and its actuation principle lays the groundwork for designing ERS.

2. 2.

   System behaviors: ERS are capable of changing shape over time, therefore, their development involves determining a specific behavior. A behavior is defined as the sequence in which the system changes shape in response to feedback parameters (Fig. 4, bottom). Depending on the project’s goals, these parameters can represent various conditions. Behaviors explored in ERS can be summarized into the following categories:

   • Design-based: shape changes are driven by design criteria such as functional, volumetric and geometrical considerations. This involves identifying how a geometrical state corresponds to a specific design objective and changes over time. For example, a system may be designed to transition from flat to 3D for deployability, from open to closed to alter spatial conditions, or from one shape to another to enable specific tasks like crawling.

   • User-based: the user(s) can directly control the system’s shape through an interface based on personal preferences. The interface can display different feedback information such as the state of the structure or the system data, to enable the user to make informed decisions.

   • Human-responsive: various sensors such as cameras, proximity and heartbeat sensors use the state of humans interacting with the structure as feedback to generate control actuation values and geometrical states.

   • Environmental-responsive: environmental conditions such as sun position, temperature, luminosity, and humidity can be used to determine the geometrical state.

   • Structural-based: this mode focuses on structural parameters such as the bending radius, tensile strength and curvature to define an optimal state based on structural optimization.

   • Robotic-based: robotic parameters, including physical sensors, are used to define global shapes. For instance, load sensors can identify optimal states of the system for control value computation.

   • Sound-based: this mode involves either generating geometrical states based on sound parameters or producing sound during state changes by actuators.

Most ERS projects develop decision-making control strategies that negotiate between these different modes/parameters. While each ERS may have a preferred mode based on the project brief, each system undergoes a characterization process where crucial parameters are stored. This allows for the creation of decision-making control strategies that consider multiple factors. For example, a system might prioritize achieving a state that fulfills certain goals while ensuring structural stability. Operating within a cyber-physical network, where digital and physical data are exchanged, ERS control logic can integrate sensor data and simulation parameters to enable complex behaviors and respond to unexpected changes of conditions.

The ERS design process consists of a selection of an elastic structural response and an actuation strategy, followed by the identification of desired behaviors. The figure illustrates this approach in two parts: the upper section displays a catalog of elastic structural responses and actuation strategies utilized in ERS design. It also includes images from ERS projects: (1) Self-Choreographing Network; (2) Tensformer; (3) PNEUmorph; (4) ELAbody; and (5) EMObot. Below, the figure presents a high-level diagram that describes the decision-making control process implemented to achieve the design of desired behaviors:

Full size image

3.2 ERS multidisciplinary design and control framework

The development of each ERS is determined by specific design intentions, yet all systems share a foundational approach and a technical framework applicable to diverse structures. The initial design phase entails identifying design intents that align with the project’s overarching goals. This process unfolds across five distinct stages (Fig. 5):

1. 1.

   Identification of design goals, parameters and evaluation criteria;

2. 2.

   Multi-state design;

3. 3.

   Robotic system;

4. 4.

   System calibration and characterization;

5. 5.

   Cyber-physical behavior.

ERS framework and its processes across the five main development stages: 1) Project Brief: Identification of design goals, parameters, and evaluation criteria; 2) Multi-state design; 3) Robotic system; 4) System calibration and characterization; 5) Cyber-physical behavior. The figure includes the following images: a) Simulation studies from Elastic Choreographies; b) Custom interface from ADOOB and WINGS; c) Calibration and Characterization analysis from ELAbody; d) Robotic System from ELAbody

Full size image

3.2.1 Identification of design goals, parameters and evaluation criteria

• 1. Design goals are determined using one of three approaches:

  Fundamental design: this category refers to research projects or parts thereof, where the main focus is on development of the robotic structural system from a fundamental perspective (Fig. 6, left). In this approach, the specific application of the system is a secondary aspect, and the development is often bottom-up and exploratory. The primary attention is on the development of an enabling technology that can be applied to different contexts that require shape-morphing lightweight materials. The approach lies in the design of a specific structural-actuation principle and its shape-morphing capabilities. This includes the implementation of methods to design, characterize and control the system. This approach is fundamental for technical and conceptual advancements of the ERS framework. It starts with the identification of an elastic structural behavior and actuation principle. The integration of the two approaches enables the generation of a shape-morphing elastic system. Once the system is designed, built and operates, its capacity to change shape is then analyzed. This phase includes the determination of design and actuation parameters in relation to the main geometrical states. The development starts with the most basic arrangements at a medium scale and continues with the progression of scale and actuation parameters, increasing gradually the level of complexity of the system. Once the system achieves the human scale, it can be used as a precedent and tested in the context of specific applications.

  

  Figure showing a selection of ERS built projects organised based on design goals. Images include: (1) LOOPS simulation studies; (2) PPS physical system and curve folding pattern; (3) WINGS crawling robot; (4) TPop deforming in response to sound; (5) WINGS architectural proposal; (6) LOOPS architectural proposal

  Full size image

  Shape-morphing behavior(s): in this scenario, the primary objective of an ERS is to achieve a specific behavior. This might be recognized as the system’s ability to self-form, attain certain shapes, undergo controlled shape transformations, respond to a specific parameter or perform a particular task (Fig. 6, middle). For example, this could involve a system designed for self-deployment, transitioning from two-dimensional to three-dimensional configurations. Alternatively, it might refer to developing a system capable of morphing into states that fulfill high-level functional requirements, such as creating volumetric spaces, enclosures, canopies, and walls. Another possibility is designing a system to perform actions such as crawling or responding to specific parameters, including user preferences, human movements or environmental conditions. This phase is integral to all ERS projects, as each system is developed to achieve a particular behavior or tested based on these capabilities.

  Application/vision: the most recent developments in ERS are built upon the foundational systems, with aims centered on extending these initial structures to realize larger systems and more ambitious architectural and design visions (Fig. 6, right). This includes the aggregation of ERS units into larger configurations to envision large-scale urban structures capable of shape shifting in response to human preferences.

• 2. Design parameters: once the main goal is identified, the design process involves the definition of a sequence of geometrical desired states and the selection of a structural-actuation principle. During this phase, high-level design and qualitative criteria are translated into geometrical, material and quantitative data. This process turns the desired sequence of states into a series of specific geometrical and volumetric parameters which inform the selection of parts, materials, actuators, geometries and the assembly logic of the shape-morphing system. The variation of these parameters will result in changes of shapes.

• 3. Evaluation criteria: The design objective of each ERS is to accomplish a certain ‘task’, which is translated into a sequence of volumetric states. Each volumetric state is described by a set of points representing significant positions on the overall geometry. The number of points selected depends on the level of accuracy needed to achieve the task. Once a state is clearly defined, the evaluation criteria incorporate actuation and physical parameters. The overarching goal for each ERS is to achieve the desired sequence of states with minimal actuation and material use. For example, if the objective of a system is to crawl, the evaluation process considers how a specific structural-actuation system can change position with the least required actuation. This principle is similarly applied to other transformations, such as the opening or closing of a building canopy, wall, or any other intended transformation. Quantitative evaluation criteria can also be extended to accommodate qualitative design goals, allowing for the addition of elements to create specific qualities such as a sense of enclosure, reduced transparency, or other aesthetic goals unique to the project. Due to the multidisciplinary nature of these projects, the evaluation criteria always involve a negotiation between qualitative and quantitative factors.

3.2.2 Multi-state design

The development of ERS begins with the identification of the structural-actuation principle that is most suitable to the project brief, based on project-specific evaluation criteria. Once the structural-actuation system is selected, the design process takes place through simulations and physical prototypes. As with other elastic systems where form emerges from material behaviors, ERS design involves a process of form-finding. However, ERS approaches differ from other traditional elastic structures because they are systems designed to achieve multiple states. Therefore, the process has to be extended to the study of multiple shapes and actuation sequences. This phase can be summarized in three steps:

1. 1.

   Numerical and analog form-finding at multiple states involves simulating the structural-actuation system using physics modeling engines and calibrating the system based on analog studies. The current ERS modeling tool used is the Kangaroo 2 physics plug-in for Grasshopper (Piker 2017). The geometry of the material system is computed by the Kangaroo solver, guided by a series of ‘goals’ components. These components represent the elements of the system and their behaviors, whether bending or tensile. Each goal possesses a strength that determines its influence on the solver. This strength can be used as hypothetical values in early design stages or calibrated for more accurate material behavior representation. Actuation principles are integrated into the simulation through custom components in Grasshopper and Kangaroo, with variable parameters. Altering parameters such as topological conditions or actuation forces change the solver’s output. This process also involves integrating project-specific Grasshopper components developed in Csharp to utilize existing Kangaroo functionalities. This allows for a parametric design process where topological and geometrical features can be easily modified, enabling the exploration of global geometrical outputs. Design studies then analyze how these variables affect the shape and state of the system. Given the difficulty in accurately representing these systems, the simulation undergoes calibration through a series of physical prototypes. These prototypes are fabricated, actuated, and analyzed at key states using digital image correlation, measurements, and RGBD cameras. This process allows for the reconstruction of the physical system’s mesh geometry and its comparison with the simulation results. Subsequently, the main numerical parameters are calibrated based on this comparison.

2. 2.

   Structural analysis of main states: Once the Kangaroo model is calibrated, the resulting meshes are input into K2 Engineering (Brandt-Olsen 2016b) for structural analysis. This involves assigning structural properties to the elements within the simulation, and incorporating real-world forces and loads. Parameters such as Young’s modulus, moment of inertia, length, cross-sectional area, and pre-tensioning values are factored in. In addition, various types of loads are applied to the model. These include passive loads such as the self-load (dead load) of the model, snow load, and lateral loads such as wind load, as well as active loads used to deform the shape, such as pressure in pneumatics or nodal forces. The structural calculation results are generated through the Kangaroo solver. K2 Engineering’s structural analysis components then utilize the mesh output from the solver to provide crucial structural information. This includes data on maximum displacement and maximum load in the system elements, which are essential for predicting the system’s stability under both dynamic and static conditions.

3. 3.

   Parametric design studies and the creation of a shape catalogue involve the categorization of main geometrical states based on structural-actuation principles, geometrical and topological parameters, and actuation sequences. These are formatted as a series of parametric definitions in Grasshopper/Kangaroo. The design catalogues compiled in this manner store meshes and geometrical analyses of the main states of different systems. Based on specific design criteria, a particular system is then chosen to be developed into a robotic system.

3.2.3 Robotic system

The development of the robotic system for ERS encompasses the formulation of a fabrication and actuation strategy for the structure. This process involves selecting appropriate materials, designing component parts, determining the assembly sequence, creating the mechanical design of the actuation system, and developing a feedback-control strategy.

ERS actuation typically employs motorized and pneumatic systems equipped with feedback capabilities. Sensors and actuators are managed by custom algorithms implemented on micro-controllers, such as Arduino or Raspberry Pi, operating within the ROS (Robot Operating System) environment (ROS 2007). ROS, a widely recognized robotics framework, provides a suite of software libraries and tools suited to versatile application development. It facilitates efficient communication and device integration, enabling the exchange of messages among different nodes (executable programs). This setup fosters a flexible and project-independent core control logic for ERS, adaptable to a variety of sensors and actuators while retaining a consistent core structure. In addition, ROS enhances the interaction between the robotic ecosystem and other software applications, as will be detailed in the subsequent section.

Over time, the fabrication process of ERS has evolved, incorporating various materials, fabrication techniques and bespoke mechanical systems.

Bending-active systems in ERS projects are constructed by connecting glass fiber and carbon fiber linear rods, typically bundled and joined with custom 3D-printed components. These linear systems can be aggregated in networks to create complex geometries. In bending-active tensile hybrid systems cables and tensile surfaces are connected to the bending rods. In these systems the actuation takes place either altering topological conditions or through variable-length cables. 

The first type uses robotic joints to alter the structure’s topology and/or anchor point conditions, with sliding and rotating joints. The second actuation method involves a motorized incremental tensioning mechanism with a spool  system controlling cable length, connecting tensile and bending elements.

The development of ERS robotic systems occurs across various scales, aiding in the understanding of how parameters change with system size.

For motorized actuation, ERS uses Dynamixels AX, XL, and MX servos (Dynamixel 2023), which offer a range of motors with built-in feedback and control modes. The accompanying software development kit (Dynamixel e-Manual 2023) allows for the use of different servo types and functionalities, while maintaining a consistent control logic.

Fabricating pneumatic systems involves complex, custom-developed strategies to create specialized pneumatic cushions. These cushions, with intricate seam patterns, are heat-sealed and reinforced with custom-cut bending plates. Various materials, including tailored composites, latex and PVC, are experimented with to achieve the desired behaviors.

Pneumatic actuation control is executed using pressure regulators, valves and pressure sensors. Components are chosen based on pressure ranges and the scale of each system, with pressures varying from 0.1 to 0.8 bar. Yet, the control logic remains consistent across all pneumatic systems, facilitating methodological evolution and functional diversity over time. This approach enables the creation of a broad range of systems at different scales.

Hybrid systems integrate previous strategies, combining incremental tensioning and pneumatic actuation into a cohesive system.

3.2.4 System calibration and characterization

Once the robotic prototypes are built and operational, each system undergoes a phase of characterization and calibration. In this stage, the primary behavior of the system is analyzed, which involves identifying key geometrical states and actuation parameters. Different actuation values are tested, and the system’s geometry is assessed. Actuation sequences are also evaluated based on the project’s aims. The most satisfactory sequences are stored and used as reference values.

In earlier ERS projects, data were collected through local sensing feedback, manual measurements, and photogrammetry to reconstruct the geometry. However, with the introduction of tracking and RGBD cameras into the sensing toolkit, geometry is now reconstructed by extracting information from these devices. This process involves capturing point cloud data through custom algorithms based on OpenCV (Bradski and Kaehler 2008), a computer vision library, and from an RGBD camera. These data are used to rebuild a mesh of the system, in Grasshopper (Associates 2023).

3.2.5 Cyber-physical control

Controlling large elastic behaviors presents significant challenges. Despite advancements in simulation technologies, there remains a notable discrepancy between simulated and actual material behavior (Soana et al. 2020). While simulations offer approximations of reality, they are limited in predicting all disturbances that might impact a continuously operating large elastic system in an uncontrolled environment (Soana et al. 2020). Moreover, adaptive architectural systems required to respond to rapidly changing conditions and balance various parameters in real time. This challenge led to the development of a cyber-physical control network where digital and physical data can be processed in real-time in order to compute control values and achieve the system's target behavior (Fig. 7).

The evolution of the cyber-physical control network over time, and applied to different ERS projects. In the center is the main structure comprising the connection between the simulation, control and interface environments and physical system. The surrounding boxes show how different modules have been integrated over time

Full size image

The fundamental structure depends on the exchange, processing and computation of data between three primary environments: simulation, user interface, and the robotic environment. This integration facilitates the incorporation of physical data into simulation and the creation of a digital twin of the robotic system. This approach enables the real-time computation of control data through custom-control logic that is based on the negotiation between digital and physical parameters. As described in previous sections, the simulation takes place through Rhino/Grasshopper/Kangaroo and the structural analysis in K2 Engineering. The user interface is developed in the Unity game engine (Unity 2005), where a digital twin of the system is visualized in addition to data from the simulation and robotic system.

This enables the development of a visualization and an interactive control space for the user. This is possible because the simulation environment (Grasshopper) can run inside Unity using the Rhino.Inside library and communicate with ROS through ROSsharp. ROSsharp consists of open-source software libraries and tools in Csharp specifically designed for communicating with ROS from. NET applications, such as Unity. Once this framework is operational the designer can build the control logic of the system, defined as behavior, and determine different modes of operation. For example, control values can be computed as a negotiation between design inputs (in the form of geometrical states), material and robotic feedback (bending radius, robotic joint load and position). Based on characterization data, crucial material and actuation parameters are stored in the network. Real-time feedback from the physical and simulation environments can monitor and process relevant parameters such as bending radius, tensile curvature, and geometrical deformations, as well as load, pressure and RGBD camera sensors data in order to compute the next desired state. Based on the project brief, other sensors and feedback can be integrated. They can integrate functional, environmental and human conditions. All these data are fed into the control logic and processed through custom algorithms (in Unity and ROS) in order to compute control values and achieve the desired state and behaviors of the system.

3.3 Overview of built ERS

Based on the aforementioned criteria, we present the evolution of selected ERS, focusing on their structural-actuation principles (Figs. 8, 9), and outline the specific behaviors explored for each system. Multiple ERS have been developed since 2018. While each project has emerged in a different context, all share the same fundamental approach. Figure 9 provides an overview and classification of presented projects.

ERS projects overview and classification—1

Full size image

ERS projects overview and classification—2

Full size image

In the first ERS, ‘Self-Choreographing Network’ (Maierhofer et al. 2019), the structural-actuation principle relied on continuous topological changes within a network of elastic beams. The bending-active rods, actuated by robotic joints, were capable of sensing and inducing both local and global deformations. Control values in this system were generated, evaluated, and updated in real-time, utilizing a custom decision-making strategy that balanced design, material and robotic parameters. The primary behaviors explored in this project were design and robotic-based.

Following this work, ELAbot was developed in 2020 (Soana et al. 2020), setting the fundamental basis for the current ERS framework and approach. Its structural-actuation principle was based on the incremental tensioning of a bending-active tensile hybrid loop. This project was key in developing and conceptualizing different behaviors. The framework introduced by this project enabled real-time simulation feedback, human movement tracking, and integration of user preferences through a custom interface. The system was tested and developed using design, structural, human and user-based behaviors.

LOOPS (2021) (Soana et al. 2022) extended the concept by exploring the aggregation of multiple ELAbot modules to develop a mobile and self-shaping system. The primary behaviors explored in LOOPS were design-based and human-responsive. For complex behaviors like crawling, LOOPS employed a machine learning-based approach to determine optimal control values for achieving the desired behaviors.

ELAbody (2021), exhibited at the Venice Biennale 2021, consisted of a network of bending-active rods. These rods were arranged to function as a bending column or exoskeleton, capable of mimicking human body movements.

Two bending-active tensile hybrid modules were incorporated on either side of the central bending elements. At the base of the structure, three robotic incremental tensioning mechanisms were integrated. These mechanisms connected the end of the central rod with the two tensile elements on the sides using variable-length cables. ELAbody utilized the same actuation system as both ELAbot and LOOPS.

In a novel approach to human–structure interaction, ELAbody was directly connected to a human body, with human arms linked to the structure, allowing the system to move in tandem. This design explored embodied relationships between humans and elastic structures, presenting potential applications in fields such as rehabilitation or performance art.

Elastic choreographies (2021) (Sabery et al. 2023) explored topologies similar to ELAbody but differed in its actuation principle. This project featured modules of bending-active tensile hybrid elements with triangular typologies, manipulated by custom robotic arms that could change the angle and position of the modules’ anchor points. It delved into multiple behaviors, with the most notable being the development of a machine learning-based environmental behavior and a structural-based behavior for achieving dynamic shading and optimal structural conditions.

From 2021, the ERS framework began to include new methods developed incorporating inflatable systems, pneumatic actuation and geometrical reconstruction techniques. Based on these advancements, new ERS projects were developed. The first project was PNEUmorph, a shape-morphing pneumatic membrane constrained by a network of variable-length cables. Through this project the feedback-based pneumatic system was developed and integrated in the cyber-physical network. This was also the first project where RGBD cameras were used to perform a geometrical reconstruction of the system. These methodologies were the base for the development of further ERS, such as WINGS, which focused on developing a new enabling technology, comprising pneumatic cushions with patterns of tailored seams reinforced by perimetral bending-active plates.

WINGS (2022) also extended this aim to the development of human-scale systems and interactive performances, where global deformations were informed by the facial expressions of the user. Inspired by this, two other projects, EMObot and A Dance of Other Bodies (ADOOB), were completed in 2023. Conceived as pneumatic hybrids, they combined bending-active plates and inflated systems with curve folding principles.

EMObot (2023) extended the geometrical and shape-changing catalogue of WINGS. This was achieved by integrating bending plates as constraining layers of the cushions to create bending shapes. In the human scale series of demonstrators, the system behavior was based on human states, based on feedback from human heartbeat, breathing rate and facial expression.

ADOOB (2023) inverted EMObot structural systems by integrating the bending layer inside the pneumatic cushion for specific deformations, namely the global bending of a leaf-shaped module. As a behavior, this project explored how pneumatic and robotic noises could be altered with sound processing techniques to generate melodies.

TPop (2022) involved a dynamic hybrid textile surface connecting bending-active pneumatic tubes, actuated with variable-length cables. TPop was designed to respond to music and deformations that were controlled by sound parameters.

In progress ERS are focusing on large scale systems, exploring the integration of new interactive methods and specific applications. This includes interactive assistive devices, light-responsive facade systems, sound-responsive building elements, a shape-morphing pavilion, and ergonomically adaptive furniture.

This paper has proposed a framework for the design and control of shape-morphing systems that display complex behaviors and multiple fields of applications. It shows how this framework seamlessly integrates methodologies and tools from architecture, engineering and soft robotics, and has been applied in the development of different ERS projects. It serves as guide for designers and engineers who wish to develop novel ERS. The ERS framework also proposes a new vision for interactive systems that offer immersive experiences for humans in an intelligent way, leveraging material properties and minimizing resource use.

While our current framework has shown promising results, further development is necessary in three main areas. First, advancements in fabrication techniques of multi-material composites are essential to enhance the efficiency, scalability, and durability of the system. Second, technical enhancements are necessary to integrate more complex decision-making processes that can learn and evolve over time based on feedback. Lastly, there is a need for a robust strategy to upscale the framework and comprehend the structural implications of large-scale changes, particularly for controlling long-term behaviors. In addition, improvements and optimization of the cyber-physical control are required to enable long-term functionality. These areas represent key avenues for future research and development, aiming to address the identified limitations and unlock the full potential of the ERS framework.

ERS presents a new vision where the design process emerges through a continuous conversation between the designer, the material and the robotic system. This can lead to the generation of sustainable, adaptive, lightweight systems that adapt to changing conditions, maximizing their performance. In the realm of architecture, ERS could enable the development of adaptive envelopes that respond to weather conditions, optimizing energy efficiency while maintaining indoor comfort. Imagine a building exterior that morphs to maximize sunlight intake during winter or create shade in the summer, without the need for external intervention. Similarly, interior spaces could transform based on occupancy or use, such as walls that expand or contract to create new rooms or change their acoustics for different events.

ERS could also lead to the creation of furniture that adjusts to the user’s body for optimal ergonomic support or changes shape to suit different uses. For example, a chair could morph into a lounge for relaxation or a desk for work, responding not only to user commands but also to sensed needs, like adjusting support in response to the user’s posture. Beyond functionality, these advancements could usher in new forms of aesthetic expression and experiences. Architectural spaces and furniture could become platforms for interactive art, changing form and appearance in response to human presence or environmental stimuli, creating an immersive and constantly evolving environment. This could redefine how humans interact with and perceive their surroundings, turning spaces and objects into partners that actively respond to and engage with their users.

The progression of ERS promises to transform not only the functionality of spaces and objects but also their aesthetic and experiential aspects. As these advancements continue, we can envision a future where architectural spaces and furniture transcend their traditional roles, evolving into interactive art forms. These environments would dynamically change form and appearance in response to human presence or environmental factors, offering an immersive and continuously evolving sensory experience.

In such a world, how humans interact with and perceive their surroundings would be fundamentally redefined. Spaces and objects could become active participants in our daily lives, responding to and engaging with users in meaningful ways. The relevance of living in robotic spaces is highlighted by objects that communicate information in an embodied manner, offering a more intuitive and natural way of interaction.

We can speculate that robotic systems will evolve from the mechanical and utilitarian entities we are accustomed to into ubiquitous elements that enhance everyday experiences. Imagine a living space where walls could subtly change colors or textures to reflect the mood or needs of the inhabitants, or furniture that not only adjusts to our physical needs but also conveys information or emotions through its form and movements.

These advancements would lead to a blending of technology, art and daily living, where the boundaries between each become increasingly fluid. The potential for ERS to enhance not just physical comfort but also emotional wellbeing and aesthetic enjoyment is vast. As these systems become more integrated into our environments, the distinction between living spaces and robotic systems would blur, leading to a new era of human-centric, responsive, empathetic architecture and design.

The possibilities for creative expression and personalization of spaces are limitless. From homes that adapt to their inhabitants’ lifestyles to public spaces that interact with visitors in engaging and informative ways, the integration of ERS into our environments could significantly enrich human experience and interaction with the built world.