Most of the current adaptive system designs are focused on the programming of the software. In this report, however, what we are really interested is more on designing entire systems rather than software progrmming, that is, including the physical body of the system, its morphology. Nicholas Negroponte [2] argues that the machine needs to have a body like us in order to think and behave like us, this leads to the discussion of how the human body contributes to intelligent behaviours? The concept of embodiment [3] is concerned with the relation between physical and information (neural, control) processes. Morphological desgin is about connecting body, brain and the environment. The creature-like characteristics and specific behaviours of a machine may both be referred to as the natural morphology of the physical body, and the sequence of activation. Morphology of the physical body is crucial to the abilities and features of the machine, e.g. the sequence of activation (choreography of movements) enable abilities such as walking, lifting or snake-like locomotion [4].
It is important to mention the co-evolution of morphology and neural substrate in biological systems and the reciprocal relationship between the two in terms of adaptive behaviour. This research will take a morphological approach to design behaviours, by means of a design project, the results will show that by developing appropriate morphology and choosing right materials we can create intelligent behaviours with very simple control methods. We will proceed as follows. First, we illustrate the role of morphology in intelligent systems. Second, we examine the relations between morphology, material and control with some introductory examples and experiments. Third, we investigate the exploration in morphology, control and materials through an own design project. Finally, we will summarise the most important insights gained from our research.
2. Scope and Limitations
There is lot of research on the morphology of sensory systems; for instance, Franceschini and coworkers [6] found that in the housefly the spacing of the facets in the compound eye is more dense toward the front of the animal to compensate motion parallax. However, this research is focused on the morphology of body parts, actuators and materiality, examine the relationships between these parameters and adaptive behaviours, the morphology of sensory is not in the main scope of this paper.
3. Role of morphology in intelligent systems
3.1 Biological systems
Researchers often use robotics to study intelligent systems, modeling many different aspects of biological systems. A new approach in the design of intelligent systems draws on inspiration from biology, instead of using traditional concepts of artificial intelligence. The role of morphology and materials in biological systems can be explained using a motion control task [8]. While approaching the glass we open the palm and applying force to wrap our fingers around the glass. The fingers will automatically adapt to the shape of the object passively [9], this is achieved by the softness of tissue in the hand and on the fingertips, not centrally controlled.
The above example shows that the morphology of body takes over some of the processing load from the brain. Processing in this sense is embodied in the physical settings of the system, the intelligence exists in the continuous interactions of an embodied agent with the real world, not in the central processing software alone, the morphology of the agent is critical to its adaptive behaviours.
3.2 Artificial systems
Pfeifer [7] argues that if we are interested in designing adaptive systems, we should aim for ‘emergence’. The term emergence is used in this context to mean being not pre-programmed. As mentioned above, behaviour cannot be reduced entirely to internal processing; it is always influenced and triggered by its interaction with the environment. Systems designed for emergence tend to be more adaptive and robust [7]. For example, a system specifying initial conditions and developmental mechanisms [10] will automatically exploit the environment to shape the agent’s final structure.
Karl Sims [10] addressed this approach in his virtually evolved creatures; he simulated a population by evolving virtual block creatures in which each creature is set to perform certain tasks such as walking or jumping. The mutation and growth contained in the code of the creatures’ virtual genes will change the morphology of the creature according to the fitness of each generation, the morphology and the behaviour of the creature emerges from its interactions with the virtual world. What is more; these two factors affect and interact with each other.
Another approach is provided by Lipson and Pollack [11], they designed the morphology to be consisted of rods to which different types of joints could be attached. The system can be parameterised according to joint type, rod length, diameter and material properties, thus the search space of possible morphologies is constrained. These studies are closely related to our own work, in which the search space is limited to the exploration of behaviours of modular kinetic structures.
Fig.1 Karl Sims – Virtually evolved creatures for walking, different body plan and weight distribution.
Fig. 2 Lipson ‘s sample robot morphology
3.3. Materiality
Improvements in flexible and programmable materials are re-shaping robotics away from typically rigid structures driven by servomotor systems, towards more fluidic and responsive mechanics inspired by nature. For instance, human arms are built of parts that are flexible to varying degrees (bones, muscles, tendons). There is a natural position for a human arm with palms facing the body, which is determined by its structural composition and material properties. If we have the back of our hands facing the body, our arm will turn back to its natural position when we let go of the force. This is not achieved by neural control but by physical constraints provided by morphology and the material properties that make the control problem much easier.
These ideas can be readily transferred into robotics. Many researchers have started building artificial muscles and used them on robots. In this project we are particular interested in the pneumatic artificial muscle (PAM), since it has many biomorphic characteristics and is very low cost. It was first developed in the 1950s under the name of McKibben Artificial Muscles [12], and they have some properties of natural muscles: they provide a natural compliance resulted by their material properties. Pneumatic artificial muscles have many advantages such as a high contraction ratio of 25% and high power to weight ratio (400:1). Using soft and lightweight actuators, we can place the weight of a central control system strategically on the body of the structure, without being adversely affected by the weight of the motor actuators.
Fig. 3 The McKibben Pneumatic Actuator, with exterior braid and inner elastic bladder
Currently, materials research in morphological design is largely focused on the end effectors (e.g. gripers, feet) with minimal investigations on the body as a whole (There are NASA’s tensegrity robot, and Havard’s soft robots exploit on this topic which will be discussed later). This research will exploit materiality with experiments of kinetic structures in which the body parts are composed of different soft and rigid materials. Soft robots have the potential to change the way we construct intelligent systems, by using highly malleable and stretchable materials, we can build robots that safely interact with human operators and reduce computation work significantly.
4. Experiments of morphology and behaviours
4.1 Golem-Kit
Current experiments with artificial systems are mostly performed with simulations; indeed, artificial systems that are able to grow physically in the real world are only at an embryonic stage (for example, in the field of nanotechnology, growth mechanisms are investigated) [13]. Pfeifer [7] argues that one way to tackle this problem to some extent is to; on the one hand, build a very good simulation that accurately models the physics of the interaction between agent and the environment, but this could be difficult and time consuming build and computationally very heavy to run. On the other hand, alternatively, have robot-building kits that enable researchers to quickly build a robot to test in the real world. In this research we have taken the latter approach of designing and producing a fast prototyping toolkit.
The aim of the toolkit is to iterate physical models quickly and reduce the computational work in both controlling the structure and simulating the real world. This toolkit includes 3 main elements: Pneumatic Artificial Muscle; Slot-in Modular Joint; and a portable autonomous pneumatic control system. The kit provides range of different joints for potentially constructing a diverse variety of structures.
Fig. 4 Six basic types of slot-in modular joints including 3d-printed connectors
Fig. 5 Combinations of basic joints can create more complex joint with multiple degrees of freedom
We developed a low-cost version of McKibben’s pneumatic actuator [12], replacing the original rubber tube with a latex , which is much cheaper while performs the role adequately. Braided cable sleeve is used as the outer woven ‘skin’ to constrain the inflation effect to linear elongation and contraction. By using very accessible and low-cost materials to build our own PAM, we were able to build and test as many possible designs of kinetic structure.
Fig. 6 Altered version of McKibben’s pneumatic artificial muscle using balloon and braided cable sleeves.
This toolkit was brought to the public in the pneumatic kinetic structure workshop in Rio de Janeiro, Brazil. The groups participating in the workshop were given the task of building kinetic structures using our toolkit, with the limit of using maximum 9 PAMs. The aim of the workshop was both to test the capacity of our toolkit to the full extent and to experiment with various possible compositions and materials of pneumatically driven kinetic structures. The experiments carried out within the workshop will be illustrated in following sections, which are divided into different categories according to the feature being assessed.
4.2 Morphology
Tad McGeer [14] showed that a simple biped mechanism is capable of walking down an incline without any actuation or control. The robot has no motors, no sensors or processing unit, its locomotion is an outcome of gravity and the mechanical properties of the structure design (e.g. leg segment lengths, mass distribution, and foot morphology). The original design had four legs to provide lateral stability; although Collins [15] constructed a biped version that simply uses the counter-swing of the arms that are attached to their opposing legs to balance the robot.
Fig.7 Collins et al., The Cornell passive dynamic walker, 2001
The Boston Dynamics team take an entirely different approach to robot construction. Their aim is to build a robot that could perform on a large number of different terrains. RHex, a Boston Dynamics robot, is a biologically inspired hexapod robot with semi-circular arc-shaped legs and only single rotatory electric motor per leg. It is the first documented autonomous legged machine to have displayed general mobility (speeds at body lengths per second) over general terrain (variations in level at body height scale) [16]. RHexs’ high mobility is achieved through its 6 legs that together can produce adequate gaits to overcome even very rough terrains. This case study demonstrates diversity in behaviours does not always require complex morphology designs. On closer investigation of this topic, experiments of locomotion closely related to morphology and material properties will be presented below.
Fig. 8 Boston Dynamics, RHex – rough terrain robot
Grasshopper robot
Design in collaboration with Marcos Bravo, Rio workshop
This design developed in the Rio workshop explores the characteristics of bamboo in kinetic structure. Bamboo dowels make use of the curved extremity of the structure aking contacting surface with ground; two strips of bamboo, each with a PAM connected on both ends, are attached at the front and rear underside of the structure, acting somewhat like limbs. When one PAM engages, the bamboo strip will bend immediately into an arc and lift the corresponding part of the structure up, and due to the arced shape of the surface-contact component, the structure swings in the opposite direction to the limb. By alternating the actuating PAM, the robot performs jumping and swinging movements.
Fig. 9 Grasshopper Robot
4.3 Materiality
4.3.1 Combination of soft and rigid
Buckminster Fuller [17] invented the concept of ‘tensegrity’, or tensional integrity, in the 1960s. Tensegrity structure consists of isolated compression members and continuous tension members, The continuous pull force is balanced by the continuous push producing a balanced integrity of opposing tension and compression. The super-rigidity is obtained from the combination of soft and rigid elements and the mechanical composition of the two elements. NASA developed their space-bound Super Ball Bot robot using the principles of tensegrity, The robot consisted of six linear motors as compression members and six spring components as tension members. When the linear motors are actuated, the robot can change shape and accomplish locomotion under specific actuation sequence, the major advantage of this tensegrity robot for space expedition missions is the great compliance in overall structure given by its material properties, it can be flattened to a certain degree to pass narrow passages and protect the robot from impacts and strikes [18]. Using tensegrity structure to design robots inspired us to carry out our own experiments into kinetic tensegrity structures to explore how morphology and materiality play in this form of structural composition.
Fig.10 Buckminster Fuller’s Tensegrity
Fig.11 NASA’s SuperBall Bot
Bamboo Tensegrity
Design in collaboration with Fernando Daguanno, Rio workshop
This six-strut tensegrity structure consists of six bamboo strips connected by flexible strings. The very simple control architecture was used, three PAMs attached to each end of each bamboo strut, to actuate movements. Due to the relatively high elasticity and bending strength of bamboo, the structure is able to contract and expand within a small range: when one of the PAMs is inflated, the bamboo strips attached to that PAM will bend under the tension force generated by PAM, when deflated, bamboo strips will bounce back to its initial state, triggering a ‘jumping’ movement of the structure. This tensegrity structure is able to perform a ‘dance’ of eight different moves with different actuation combinations of the three PAMs.
Fig. 12 Actuated Bamboo Tensegrity Robot in jumping motion
4.3.2 Soft robot
A team from Harvard’s Wyss Institute, Harvard’s SEAS, and MIT built an autonomous origami robot that consists of a single composite sheet composed of flexible circuit board sandwiched between layers of standard letter-sized paper and pre-stretched polystyrene plastic, a small battery pack and pair of motors mounted provide power for transformation, programmed to fold itself into a complex shape and achieve locomotion autonomously. The spiky limbs are a result of form following function, the sharp angles were a result of the many folds on the sheet material, the feet are large in order to contain long folds, which provide extra torque for the robot to lift itself during assembly [19].
Fig.13 Harvard university’s origami robot
If we take a closer inspection of soft materials used in robotics, we can find designs that are able to transform and deform using extremely soft materials and are free of motors, such as pneumatic silicon robots. By carefully designing the morphology such structure can perform very delicate tasks such as locomotion through tight space or gripping fragile objects, e.g. the Soft silicon robot (Fig. 14) and soft robotic gripper (Fig.15) designed by GMW Group from Harvard University, the shape of the soft silicon robot is a flat piece of star fish – with linear arrays of air chambers in its ‘body’ and ‘limbs’. Sequential inflation and deflation of the ‘limbs’ allows locomotion of the robot, when it encounters a barrier (a glass pane), deflation of the ‘body’ can flatten the robot down to fit the height of space underneath the barrier, and a different sequence of actuation of the ‘limbs’ allow the robot to crawl through the barrier with a kind of undulatory motion.
Fig.14 GMW Group – Soft Silicon Robot
Fig.15 GMW Group, Soft Robotic Gripper,
The examples of soft robots proof that cleverly exploited morphology and material alone can provide the system key abilities to perform certain tasks, and significantly reduce the programming of software and the burden of motors in the overall system. As part of the research, an experiment of deployable paper origami structure with PAM actuators is carried out to test the behaviours of soft foldable and deployable structure.
Origami robot
Design in collaboration with Helena Porto, Rio workshop
This experiment with a paper origami robot investigates behaviours in foldable and deployable structures. A sheet of paper origami is actuated by 9 PAMs attached to 18 vertices of the folds on one side of the origami structure. The 9 PAMs are divided into 3 groups, each group consists of 3 PAMs crossing each other with 1 PAM running along the width of origami and 2 PAMs running diagonally, the groups are evenly spaced out along the length of origami in a linear array. When activated, the structure compromises the responsiveness of pneumatic actuators, and generates continuous and smooth motions between folded states along the pre-determined folding creases, meaning that the robot displays a worm-like biomorphic characteristic, but is not yet able to perform locomotion like a real worm, due to little friction between the origami and floor surface and the absence of ‘foot’ components, moreover, a better actuator arrangement and sequence of actuation are to be explored. To conclude, this experiment shows the material properties (flexible, deformable, twisting) of paper origami can interact with the movement of PAM, as the twisting of each paper facet varies every time due to complex material dynamics, and the chain reactions transmit through the structure, this leads to slightly different deployment motions every time. Nevertheless, by accurately positioning the actuators on a soft surface grid, these differences could be undermined, whilst maintaining the emergent of behaviours of the origami structure.
Fig. 16 Three stages of the origami deploying under PAM actuation
4.4 Control system
To briefly summerise the case studies and experiments, structure exploited in morphology and materiality does not require full actuation at all time compared with conventional rigid robots. We can minimise the level of actuations and still accomplish the task, and reduce the computation required. The interaction between morphology, material properties and environment (friction, surface terrain), and control (sequence, amplitude, phase) can be exploited for simplifying the control system. For example, in the case of bamboo tensegrity robot, the phase between the bending movement of the 6 bamboo struts can be varied to change the amplitude of movement.
5. Design Thesis
The project proposed a pneumatic kinetic structure featuring low-cost pneumatic artificial muscles, modular construction and lightweight structure. The project is a research tool to explore possible morphologies of locomotive structure, investigate the interplay between the morphology, materiality and control organization, and the influence they have on behaviours of such structure. The project is developed using the pneumatic prototyping toolkit we developed to iterate physical models without simulations. The project will be analysed in three categories of morphology, materiality and control organization separately, bearing in mind the close relationships between them, we will review these parameters as a complete system eventually.
5.1 Morphology
First prototype
In design of the first prototype we drew inspiration from the concept of tensegrity structures with alternations to enable the robot to walk. The structure consists of 2 mirrored interlocking tetrahedrons constructed by 12 aluminum
tubes, the upper upside down tetrahedron with the vertice pointing down to the ground becomes a leg, the upright tetrahedron with its vertice connected with the 3 other vertices of the upside down tetrahedron through springs, is the supporting structure of the overall system. 3 PAMs are vertically attached to the perimeter vertices of the two tetrahedrons, 1 PAM attached to the top vertice of both tetrahedrons. The result of this structure is satisfying for an early stage, the robot is able to perform effective locomotion. Based on observation of the redundant elements in the structure (e.g. the bottom tetrahedron) ，we decided to evolve it into a simpler modular form for the next generation.
Second prototype
Testing the first prototype it was found that enabling the robot to walk required two types of different functioned structures; one is dynamic which performs as limb, the other is rigid and solid as supporting structure. Building on this knowledge, the second prototype is an attempt to simplify the structure as a module in larger modular system, by re-composing the structure in which the two tetrahedrons interlock like chain, and maintained the movement range of the ‘leg’ same as the first generation.
Third prototype
In the third prototype we kept the same base tetrahedron and added more legs and joints to allow larger and more complex movements, but with as few PAMs as possible. However, while the result was informative, it was not entirely satisfactory. The morphology of the tripod-like prototype is not capable of walking like insects or mammals; instead, when one of the legs starts moving, the structure becomes unstable and unbalanced. This instability is mainly due to its proportional leg segment length verses the overall structure. In the absence of balancing mechanism, the structure fails to lift one leg but keep the rest of the body stable. By adding wide, arc-shaped feet like the Passive Dynamic Walker, it prevented the structure from falling over, but also reduced the movement range of the legs in exchange. This is an interesting experiment of biomorphic morphology. However, such structure is not suitable for modular systems, as it cannot be arrayed to form larger structures due to its morphology.
Final design
Learning from the previous attempts to make the robot, we decided no to adopt the biomorphic forms used in the third prototype. In the composition of human muscle and tendon system, the walking movement demands the many different muscles working together, but from testing the prototypes is became clear that the equivalent number of PAMs would require an unsustainable number of PAM control devices. Instead, in order to achieve a very compact and efficient kinetic structure, the final design turned to simple geometric and modular forms of structure. As a consequence, we successfully arrived at a final design of an autonomous walking robot. The main body of the robot is a cubic frame, with 4 contractible legs that are derived from the second prototype, located at 4 vertices of the cube, 2 at the top opposing vertices, 2 at the bottom opposing vertices in mirrored positions with the 2 top vertices. The structure is able to transform from a cube to a 4-legged creature, the method to make this structure locomotive is to shift its centre of gravity by sequential actuations of the legs, which will be further explained in the control section.
Fig. 17 Prototype 1-3 & Final design
5.2 Control
First prototype
The walking cycle of the first prototype prototype can be explained as the following; When the central PAM is inflated and hence contracts upwards, the ‘leg’ tetrahedron lifts up from the ground and remain perpendicular to the ground surface, When the central PAM is deflated and all the perimeter PAMs inflated, the ‘leg’ will be pulled down to push the ground. If the central PAM and one of the perimeter PAM were actuated at the same time, the ‘leg’ tetrahedron will lift and point towards one of three mechanically possible directions. The structural composition of the 4 PAMs provides a range of five different movements to accomplish a walk in three different directions. At this stage however, the walking movement heavily relies on control system and less on morphology. Nevertheless, this prototype proofs that the combination of soft and rigid elements is an effective approach to locomotive structures.
Second prototype
In order to minimise the size of actuator required to perform the same movement range as the first prototype, sliding bar mechanism is introduced in this prototype as a new type of joint configuration. It is a more efficient actuation and control mechanism as it utilises the geometric principle of triangles to achieve maximum displacement. We continued to use the sliding bar mechanism in the later generations.
Third prototype
We used 6 PAMs to control this 3-legged robot, each leg demands 2 PAMs to lift and bring the leg inward. The key disadvantage in this control system is that there is no mechanism to make the legs stepping forward, this requires additional number of individually controlled PAMs which will significantly increase the complexity of control, this is contrary to the purpose of our research, which exploiting morphology and material to simplify control, this is the main reason for us to discard this type of morphology.
Final Design
The control has been significantly reduced in this design, although it has 4 legs with 3 PAMs controlling each of them, the actuation sequence is very simple: each leg can stretch out with 3 PAMs pressurised at the same time, which means 1 solenoid valve can control 3 PAMs to accomplish this task. When one of the bottom legs stretches out, the structure is able to roll over in the opposite direction to the leg. Furthermore, in the event of the structure rolling over, stretching out a top leg in the same direction ensures that the structure regains balance and a steady state, which implies a half-cycle of rolling movement can be accomplished by controlling 1 solenoid valve alone.
Fig. 17 Actuation sequence of PAMs in each prototype
5.3 Materiality
So far we have used rigid rod materials to constrain the search space of locomotive morphologies as the first step, focused on exploring different morphologies using the same material in the prototype developments. However, experiments of different materials from the workshop showed us a rich selection of materials available to be tested on our final design. At the moment, what we can predict is that the use of softer materials on our final structure will help smoothen the rolling movements and add more control to it, so the movements of the structure can become safer and more friendly to people in pragmatic and perceptional aspects.
6. Future development
In order to achieve autonomous locomotion, the robot needs to be able to sense if it has rolled onto its back and update its limb movements accordingly. Hence a gravity sensor is required – knowing the updated status of top and bottom, or up and down, the control program can choose which legs to actuate for the next walking movement based on the updated position. This will incorporate sensory – actuator morphology that is yet to be exploited.
Referring to the non-homogeneous morphology of the retinas of insects and mammals, the spacing at the center is more dense than on the periphery, compared with standard cameras, which have evenly distributed light-sensitive cells. These arrangements allow animals to process visual signals rapidly and it is the morphology of retina does a lot of preprocessing. This parallel processing ability is crucial to achieving real-time reactions. Sensory morphology has to be exploited in order to do this processing. In the next stage of my research, I will explore sensory morphology in relation to real-time interactions with the environment into account that we are able to understand the role of morphology in intelligent behaviour.
7. Conclusions
This thesis has tried to discuss the results from many different kinds of case studies and experiments in the field of robotics kinetic structure, in the context of morphology and material. First, the examples have shown it is important to exploit of right morphology and materiality in order to simplify control system and achieve energy efficient movements. Second, there is a correlation or balance: the better the exploitation of the dynamics in morphology and material, the simpler the control. Finally, it is evident from the studies in this research that the interactions between morphology, material and control system are crucial to achieving real-time emergent behaviours.
There are researches that discuss how cognition can be built from the physical configurations or how it emerges from body morphology, materials properties, and the interactions of the agent with the world. We can see how the morphology, material properties and the neural processing or control support each other in development of the entire system. The things we learnt from the studies of morphology, material and control teaches us how to learn from biological systems and how to build better kinetic structures and robots.
References:
Pfeifer, R. (2000). On the role of morphology and materials in adaptive behaviour. From animals to animats, 6, 23-32.
Nicholas Negroponte, (1975) Soft architecture machines. The MIT Press, Cambridge, MA.
Andy Clark & David J. Chalmers (1998) The Extended Mind, ANALYSIS 58: 1: 1998 p.7-19
Flavia Ghirotto Santos, 2015, noMad | A Behavioural Assembly System
Hillel J. Chiela, Randall D. Beerb (1997) The brain has a body: adaptive behaviour emerges from interactions of nervous system, body and environment, Trends in Neurosciences, 1 December 1997, Volume 20, Issue 12, Pages 553–557, Cell Press,
Franceschini, N., & Changeux, J.-P. (1997). Repetitive scanning in the fly compound eye. In N. Elsner, & H. Waessle (Eds.), Twenty-Fifth Goettingen Neurobiology Conference.Go ¨ttingen: Thieme.
Pfeifer, R., Iida, F., & Bongard, J. (2005). New robotics: Design principles for intelligent systems. Artificial life, 11(1-2), 99-120.
Pfeifer, R., Iida, F., & Lungarella, M. (2014). Cognition from the bottom up: on biological inspiration, body morphology, and soft materials. Trends in cognitive sciences, 18(8), 404-413.
Bicchi, A. et al. (2011) Modelling natural and artificial hands with synergies. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 366, 3153–3161
Sims, K. (1994, July). Evolving virtual creatures. In Proceedings of the 21st annual conference on Computer graphics and interactive techniques (pp. 15-22). ACM.
Pollack, J. B., & Lipson, H. (2000). The GOLEM project: Evolving hardware bodies and brains. In Evolvable Hardware, 2000. Proceedings. The Second NASA/DoD Workshop on (pp. 37-42). IEEE.
Chou, C.P., and Hannaford, B. (1997). Study of human forearm posture maintenance with a physiologically based robotic arm and spinal level neural controller. Biological Cybernetics, 76, 285-298.
Zerda, A. S., & Lesser, A. J. (2001). Intercalated clay nanocomposites: morphology, mechanics, and fracture behaviour. Journal of Polymer Science Part B: Polymer Physics, 39(11), 1137-1146.
Tad McGeer (April 1990). “Passive dynamic walking“. International Journal of Robotics Research.
Collins, S. H., Wisse, M., & Ruina, A. (2001). A three-dimensional passive-dynamic walking robot with two legs and knees. The International Journal of Robotics Research, 20(7), 607-615.
Altendorfer, Richard, et al. “RHex: a biologically inspired hexapod runner.” Autonomous Robots3 (2001): 207-213.
Richard, Buckminster Fuller. “Tensile-integrity structures.” S. Patent No. 3,063,521. 13 Nov. 1962.
Sabelhaus, A. P., Bruce, J., Caluwaerts, K., Chen, Y., Lu, D., Liu, Y., … & Agogino, A. M. (2014). Hardware design and testing of SUPERball, a modular tensegrity robot.
Felton, S., Tolley, M., Demaine, E., Rus, D., & Wood, R. (2014). A method for building self-folding machines. Science, 345(6197), 644-646.
Illustrations:
Sims, K. (1994, July). Evolving virtual creatures. In Proceedings of the 21st annual conference on Computer graphics and interactive techniques (pp. 15-22). ACM.
Pollack, J. B., & Lipson, H. (2000). The GOLEM project: Evolving hardware bodies and brains. In Evolvable Hardware, 2000. Proceedings. The Second NASA/DoD Workshop on (pp. 37-42). IEEE.
Chou, C.P., and Hannaford, B. (1997). Study of human forearm posture maintenance with a physiologically based robotic arm and spinal level neural controller. Biological Cybernetics, 76, 285-298.
Image by Xiaoxuan Zhou
Image by Xiaoxuan Zhou
Image by Xiaoxuan Zhou
http://www.eucognition.org/index.php?page=passive-dynamic-walkers [Accessed: 24 June 2016]
http://www.bostondynamics.com/robot_rhex.html [Accessed: 24 June 2016]
Photograph by Xiaoxuan Zhou
http://www.pbs.org/wnet/americanmasters/r-buckminster-fuller-about-r-buckminster-fuller/599/ [Accessed: 24 June 2016]
http://www.wired.com/2014/02/nasas-super-ball-bot-revolutionize-robotics/ [Accessed: 24 June 2016]
Photograph by Xiaoxuan Zhou
http://wyss.harvard.edu/viewpressrelease/162/robot-folds-itself-up-and-walks-away [Accessed: 24 June 2016]
http://www.robaid.com/bionics/harvard-soft-bodied-robots-inspired-by-squid-and-