Compared to the traditional approach of the study of the control or the neural system itself in robotics, artificial intelligence and neuroscience, there has been increasing interest in the notion of embodiment not only in robotics and artificial intelligence, but also in neuroscience, psychology, and philosophy. This paper will explore the role of morphology in intelligent behaviors, in the context of kinetic and robotic design. On closer inspection, researches have shown that control depends crucially on morphology (shape of body and motor system, sensor shape and placement on agent) (Pfeifer, 2016). This research will be a testing ground of morphology, examine the key parameters of physical settings that contribute to adaptive behaviors, and the relations between control, behavior, and morphology, by means of case studies and own design project. The results illustrate complex adaptive behaviors emerge largely from the physical settings of the system; programming of the software can be traded off by morphological design to various extents.
Keywords: morphology, kinetic structure, control, intelligent behavior
Much of the current work on designing adaptive systems is focused on the programming of the software. However, what we are really interested in is not just the programming aspects, but rather designing entire systems, that is, including the physical body of the system, its morphology. Nicholas Negroponte 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 does our physical body contributes to intelligent behaviors? The concept of embodiment is concerned with the relation between physical and information (neural, control) processes. Morphological computation is about connecting body, brain and environment. The creature-like characteristics and specific behavior of machine are both referred to natural morphology of physical body, and the sequence of activation. Morphology of physical body is crucial to the abilities and features of machine, e.g. the sequence of activation (choreography of movements) enable abilities such as walking, lifting arms or snake-like locomotion (F. G. Santos, 2015).
It is important to mention the co-evolution of morphology and neural substrate in biological systems, which indicates close relationship between the two in terms of adaptive behavior; therefore, one can significantly influence the other. This research will take a morphological approach to design behaviors, by means of an own design project, the results will show that by evolving appropriate morphology and choosing right materials we can create intelligent behaviors with very simple control.
Scope and Limitations
There are many studies on the morphology of sensory system, for instance, Franceschini and coworkers found that in the housefly the spacing of the facets in the compound eye is more dense toward the front of the animal. Franceschini and colleagues built fully analog robots exploiting this non-homogeneous morphological arrangement. It can be shown that this arrangement, in a sense, compensates for the phenomenon of motion parallax. However, this research is focused on the morphology of body parts, actuators and materiality, examine the relationships between these parameters and intelligent behaviors, the morphology of sensory is not in the scope of this research. We will proceed as follows. First we illustrate the role of morphology in intelligent systems; with some introductory examples to illustrate the key factors contribute to morphology in kinetic structure. Second, we investigate the relation between behaviors, control and materials through an own design project. Finally, we summarize some of the insights gained.
2. Role of morphology in intelligent systems
2.1 Biological system
embodied intelligence, intelligence exists in the continuous interactions of an embodied agent with the real world, morphology of the agent is a key influence on the interactions/behaviors.
Researchers have using robotics to study intelligent systems, modeling some aspects of biological systems, such as how ants find their way back to the nest after finding food, how a dog catches a Frisbee while running, how rats navigate in a maze, or how people recongnise a face in a crowd. New approach to the design of intelligent systems, in contrast to the traditional terminology of artificial intelligence and robotics, draws inspiration from biology.
The role of morphology and materials in biological system can be explained in a lightly more complex motion-control task: grasping a glass. First, you look at the glass, then you reach for it; while approaching the glass you open the hand (pre-shaping) and finally you wrap your fingers around it by applying a certain force to the fingers. Note that by applying force, the fingers will automatically adapt to the shape of the object â€“ you do not need to know its exact shape . Moreover, because the tissue in the hand an on the fingertips is soft, it will also adapt to the shape of the object passively, not centrally controlled â€“ it is only the softness of the material that accounts for this. Lastly, because the skin is always a little humid, it has the right frictional properties. Imagine having to grasp the glass with metal thimbles on all of your fingers; it would be next to impossible â€“ a nice illustration of what materials do for us.
2.2 Artificial system
biologically inspired robotics; relationships between programming, morphology and behavior. Morphology is a crucial part of the engineering of the intelligent system.
If we are interested in designing adaptive systems,
we should aim for emergence. The term emergence is controversial, but we use it in a very pragmatic way, in the sense of not being preprogrammed. When designing for emergence, the final structure of the agent is the result of the history of its interaction with theâ€”simulated or real worldâ€” environment. Strictly speaking, behavior is always emergent, as it cannot be reduced to internal mechanism only; it is always the result of a system-environment interaction. In this sense, emergence is not an all-or-nothing phenomenon, but a matter of degree: the further removed from the actual behavior the designer commitments are made, the more we call the resulting behavior emergent. Systems designed for emergence tend to be more adaptive and robust. For example, a system specifying initial conditions and developmental mechanisms will automatically exploit the environment to shape the agentâ€™s final structure. Another example from locomotion (see below) is the exploitation of the intrinsic material properties of an agent: If the spring-like properties of the muscles are exploited, the details of the trajectories of the joints are emergent and need not be controlled.
Braitenberg addressed similar approach in cybernetics with his hypothetical vehicles, he used a range of different vehicle models to demonstrate the concept of connectionism, in which there is no conventional central processing unit or memory storage, the load of memory was placed in the numerous parallel distributed processing feedback loops composed by different sensors and feedback routes, which practice the representation of information, and make decisions on the vehicleâ€™s action in respond to stimuli, intelligence and characters emerge from the embodiment of information processing.
Fig1. Valentino Braitenberg. (ed. 1986, c1984) Vehicles : experiments in synthetic psychology
Karl Sims â€“virtually evolved creatures â€“ different body plan & weight distribution ,no material, virtual grid world, no friction, etc.
Most robot arms available today work with rigid materials and electric motors. Natural arms, by contrast, are built of muscles, tendons, ligaments, and bones, materials that are non-rigid to varying degrees. All of these materials have their own intrinsic properties, such as mass, stiffness, elasticity, viscosity, temporal characteristics, damping, and contraction ratio, to mention but a few. These properties are all exploited in interesting ways in natural systems. For example, there is a natural position for a human arm, which is determined by its anatomy and by these properties. Reaching for and grasping an object like a cup with the right hand is normally done with the palm facing left, but could also be done â€“ with considerable additional effort â€“ the other way around. Assume now that the palm of your right hand is facing right and you let go. Your arm will immediately turn back to its natural position. This is not achieved by neural control but by the properties of the muscle â€“ tendon system. The system acts like a spring â€“ the more you stretch it, the more force you have to apply, and if you let go, the spring returns to its resting position. Also, the human arm exhibits intrinsic damping. Normally, reaching equilibrium position and damping is conceived of in terms of electronic (or neural) control, whereas in this case it is achieved (mostly) through the material properties. Or, put differently, the morphology (anatomy) and the materials provide physical constraints that make the control problem much easier â€“ at least for the standard kinds of movements. The main task of the brain, if you like, is to set the material properties of the muscles, the spring constants. Once these constraints are given, the control task is much simpler.
These ideas can be transferred to robots. Many researchers have started building artificial muscles and used them on robots. Another highly desirable property that one gets for free if using the right kinds of artificial muscles is passive compliance: If an arm, for example, encounters resistance, it will yield elastically rather than pushing harder. In the case of the pneumatic actuators this is due to the elastic properties of the rubber tubes.
Currently there are not enough studies on the material aspect in relation to morphology in the field of kinetic structure (for example, in the field of nanotechnology, growth mechanisms are investigated), current researches are mainly focused on the materiality of end effectors, e.g. gripers, feet, but not many studies of the impact of materials on entire body. This research will exploit materiality with experiments of kinetic structures in which the body parts are made of different soft and rigid materials. Soft robots have the potential to change the way we construct intelligent systems, by using highly deformable and stretchable materials, we can build robots that safely interact with human operators.
Reference : project octopus, Harvard soft robotics
3. Experiments of morphology and behaviors
Whenever we have an embodied system, through the embodiment itself, all aspects of an agent-sensors, actuators, limbs, the neural system â€“ are always highly connected: Changes to one component will potentially affect every other component. From this perspective we should never treat each of the components separately. However, for the purpose of investigation and writing, we must isolate the components, but at the same time we must not forget to view everything in the context of the complete agent. Having said that, we now proceed with a few case studies, first focusing on the morphology side, then the material side, the control system (location of sensors and actuators) and finally the integration of them all in the final design project.
At the moment we are confined to simulation; the experiments with artificial systems that can grow physically in the real world are only in their very initial sates (for example, in the field of nanotechnology, growth mechanisms are investigated). One way to get around this problem, at least to some extent, is on the one hand to have a good simulator that modes the physics of the environment, of the evolved individual, and of its interactions with the real world (e.g., gravity, impact, friction), and on the other to have robot-building kits that enable the researchers to quickly build a robot to test some individuals in the real world, using the evolved creatures as the blueprint.
In order to iterate physical models quickly and reduce the computational load of simulating the real world, we developed a fast prototyping toolkit. This toolkit includes 3 main parts: Pneumatic Artificial Muscle, Slot-in Modular Joint, and a portable autonomous pneumatic control system. The kit provides range of different joints for constructing a large variety of structures.
Fig. 10 (Left) Slot-in modular joints kit (including push-in connectors)
Fig. 11 (Right) Types of 2-degree-of-freedom joints
In this project we are particular interested in biologically inspired soft actuators that have similar properties to natural muscles, such as the pneumatic artificial muscle (PAM), it was first developed in the 1950s under the name of McKibben Artificial Muscles (Chou and Hannaford, 1997), 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 the high contraction ratio, the speed of contraction, the high force they can produce. Using soft and lightweight actuators, we can place the weight of central control system strategically on the body of the structure, without being affected by the weight of motors actuators. The major disadvantage of pneumatic actuators is that the level of precision in control is very difficult to achieve due to the material compliance of pneumatic actuators.
Fig. 8 The McKibben Pneumatic Actuator, with exterior braid and inner elastic bladder
We developed a low-cost version of McKibbenâ€™s pneumatic actuator, using balloon as the retracting/contracting â€˜muscle â€™, braided cable sleeve as the outer woven â€˜skinâ€™ to constrain the inflation effect to linear elongation. By using very accessible and low-cost materials to build our own PAM, we are able to build and test as many possible designs of kinetic structure.
This toolkit was brought to the public in the â€˜Constructing Ecologiesâ€™ workshop in Rio de Janeiro, Brazil. The groups participated in the workshop were given the task to build kinetic structures using our toolkit, with the limit of using maximum 8 PAMs. In previous prototypes we focused on testing the behaviors of different structures with same material, this workshop provided us opportunities to experiment with various materials and designs and observe the behaviors exhibited by those structures.
3.2 Morphology of Form
- How limb morphology affects locomotion in robotics, demonstration of applying different foot morphologies on same body structure in own work
Tad McGeer showed that a simple biped mechanism is capable of walking down an incline without any actuation or control (1990). The robot has no motors, no sensors or processing unit. Its locoÂmotion is an outcome of gravity and the mechanical properties of the walker (e.g. leg segment lengths, mass distribution, and foot morphology). The original design had four legs to provide lateral stability; Collins (2001) has constructed a biped version which uses a counter-swing of the arms that are attached to their opposing legs to balance the robot.
This kind of walking is very energy efficient, and there is an intrinsic naturalness to it. However, its ecological niche (i.e. the environment in which the robot is capable of operating) is extremely narrow: It consists only of inclines of certain angles. Energy efficiency is achieved because the leg movements are entirely passive, driven only by gravity in a pendulum-like manner. To make this work, a lot of attention was devoted to morphology and materials. For example, the robot is equipped with wide feet of a particular shape to constrain lateral motion, soft heels to reduce instability at heel strike, counter-swinging arms to negate yaw induced by leg swinging, and lateral-swingign arms to stablise side-to-side lean.
In terms of the design principles, this case study illustrates the principles of cheap design and ecological balance. The passive dynamic walker fully exploits the fact that it is always put on inclines that provide its energy source and generates the proper dynamics for walking. Loosely speaking, we can also say that the control task, the neural processing, is taken over by having the proper morphology and the right materials. In fact, the neural processing reduces to zero. At the same time, energy efficiency is achieved. However, if anything is changed (e.g., the angle of the incline), the agent ceases to function. This is the tradeoff of cheap design. In or der to make it adaptive, we would have to add redundancy. There is no contradiction between cheap design and redundancy: Even highly redundant systems such as humans exploit the givens of an ecological niche (e.g., gravity, friction, motion parallax)
Fig.4 (Left) Collins et al., The Cornell passive dynamic walker, 2001
Fig. 5 (Right) Boston Dynamics, RHex - rough terrain robot
A different approach has been taken by the Boston Dynamics team. Where the goal was to have a robot that could perform on a large number of different terrains. Implying that its ecological niche is larger than that of the passive dynamic walker. RHex is a biologically inspired hexapod robot with semi 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) (R. Altendorfer, 2001). Its high mobility is achieved mainly by its 6 semi arc shaped legs that produce specialized gaits to overcome rough terrains.
This case study demonstrates that the complicated and sophisticated designs typically employed in mechatronics are not always required in order to achieve behavioral diversity. Rather, the interaction of body dynamics (as determined by materials and mass distribution), environment (friction, shape of the ground), and control (amplitude, frequency) can be exploited for control purposes. There have been a number of locomotion studies concerned with the use of material properties
Grasshopper robot - Marcos Bravo
This model pushes the materiality of bamboo further to explore its characteristic in kinetic structure, the wooden boat-shaped structure utilises curve the surface-contact components, two strips of bamboo each with a PAM connected on both ends are attached at the front and rear bottom of the structure, acting like limbs. When one PAM is actuated, the bamboo strip will bend immediately into an arc and lift the corresponding part of the structure up, and because of the arc shape of the surface-contact component, the structure swings towards the opposite direction of where the limb is, when the PAM is deflated, the structure swings back to the other direction. By alternating the actuating PAM, the robot performs jumping and swinging movements.
Fig. 13 Grasshopper Robot, locomotion trigger by the bending of front & rear limbs
In conclusion, as suggested by the principle of ecological balance, there is a kind of tradeoff or balance: The better the exploitation of the dynamics, the simpler the control, and the less neural processing will be required.
3.3.1 Combination of soft and rigid
Buckminster Fuller invented tensegrity, or tensional integrity in the 1960s. Tensegrity structure consists of isolated compression members and continuous tension members, The continuous pull is balanced by the discontinuous 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 structural composition of the two elements. In general, from this case we find that even swarms of two opposite elements can generate intelligence and emergent behavior if they are organized in a smart way.
Fig.6 Buckminster Fullerâ€™s Tensegrity
Bamboo Tensegrity - Fernando Daguanno
This 6-struts tensegrity structure consists of 6 bamboo strips connected by strings. A very simple control architecture of 3 PAMs attached to all ends of the 6 bamboo struts is used to actuate movements. Due to the relatively high elasticity (16,170N ) and bending strength (20.27N) 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 and hence trigger a â€˜jumpingâ€™ movement of the structure. This tensegrity structure is able to perform a dance of 8 moves with different actuation combinations of the 3 PAMs.
Fig. 13 Actuated Bamboo Tensegrity Robot in jumping motion
3.3.2 Soft robot
a. Influence on the choice of soft actuator used in own design work
b. Analysis of soft materials (e.g. paper, bamboo) used in own design project.
Other soft structures can also obtain rigidity by various methods, such as inflating air into silicon chambers, by designing the morphology such structure can accomplish specific tasks, e.g. the Soft robotic gripper designed by GMW Group from Harvard University, a piece of star fish - shaped soft silicon with air chambers throughout, by inflating the structure, â€˜fingersâ€™ of the gripper starts to expand and fold inward, until the incurve is big enough to hold the object tight and the amount of air inflated is enough to provide required rigidity.
The control is extremely simple-the robot is virtually brainless. The reason this works is that the dynamics, given by its morphology and its materials (elastic, spring-like materials, surface properties of the feet), is exploited in clever ways.
Fig.7 GMW Group, Soft Robotic Gripper, Harvard University
Origami robot- Helena Porto, Nano Lab
This experiment with paper origami robot investigates behaviors in foldable and deployable structure, origami is a flexible geometry which can be deformed into many possibilities by simply changing the position of the actuators or activating specific points. 9 PAMs are arranged in parallel positions on one side of the origami surface and attached on 18 points, with 3 PAMs in a group and the groups are evenly spaced out. When actuated, the structure compromises the promptness of pneumatic actuators, and generates continuous and smooth motions between folded states along the pre-determined folding creases, the robot displays a worm-like biomorphic characteristic. This experiment shows significant morphological evidence in the combination of soft actuators and soft material/structure, the level of control gained is not as much as in more rigid structures, the compliance of material and structure will cause a difference in movement every time. Nevertheless, by accurately positioning the actuators on a gridded soft surface, the difference could be undermined, with the emergent property of behaviors retained. Deformation of paper
Fig. 13 Three stages of the origami deploying under PAM actuation
3.4 Control system
Muscles â€“ control from materials
The passive dynamic walker had no actuation. However, the energy efficiency of this approach can be preserved on incrementally adding actuation. This has been done by Martin Wisse and his colleagues at Delft University in Holland  in the construction of the almost passive dynamic walking robot Mike. Mike uses pneumatic actuators, which are a kind of artificial muscle: It consists of a rubber tube embedded in a fabric and contracts when air pressure is applied, which employ elastic materials (in Stumpy and in the springs of the quadruped Puppy).
Soft robotics does not require full actuation at all time, compared with classic rigid robotics. We can minimize the actuation and still accomplish the task, and reduce the computation required.
4. Design Thesis
Golem proposes a pneumatically driven kinetic structure utilizing low-cost pneumatic actuator and fast, lightweight construction. The aim of the project is to explore possibilities in kinetic structure focusing on the study of geometry and mechanical structure in transformation, and to observe the evolution of morphologies and sequence of activations (control).
In the first stage of our design thesis, we experimented with same material but different structural compositions. Our first prototype design derives from tensegrity structures, we made a cubic tensegrity with 3 struts and 12 air muscles, the struts lean against each other to create central support of the cube, we used air muscle as tension members. We found the struts are actually forming two form-shifting tetrahedrons that are horizontally mirrored with each other. We changed them into two separate tetrahedrons interlocking with each other, the upper upside down tetrahedron with the vertex pointing down to the ground becomes a leg-like structure, the upright tetrahedron with its vertex connected with the 3 other vertexes of the upside down tetrahedron through strings, is the supporting structure of the overall mechanism. 3 Pneumatic Artificial Muscles are vertically attached to vertexes of the bases of the two tetrahedrons, 1 PAM attached to the top vertexes of both tetrahedrons.
The walking cycle: When the central PAM is inflated and hence contracts upwards, the â€˜legâ€™ 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 â€˜legâ€™ 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 â€˜legâ€™ tetrahedron will be lifted and pointing towards 1 of the 3 directions. The composition of the 4 PAMs provides a range of 5 different movements to accomplish a walk in 3 different directions. At this stage, the walking movement heavily relies on control system and less on morphology, although rubber footings were used to increase the friction between the tips of contacting points and the ground surface, it is mainly the combination of soft and rigid elements that provides a level of freedom for possible movements.
From the 1st prototype we learnt that walking movement requires two types of different functioned structures, one is dynamic which performs as limb, the other is rigid and solid as supporting structure. For the 2nd prototype we tried to simplify the structure as a module in larger modular system, we re-composed the structure in which the two tetrahedrons interlock like chain, sliding bars are used in this prototype to introduce new way of controlling mechanisms using PAM.
In the 3rd prototype we kept the same base tetrahedron and added more legs and joints to allow bigger and more complex movements with less PAM as possible, however the results are not satisfying enough, the morphology of the tripod-like prototype 4 is not eligible for walking like insects or mammals, when one of the legs starts moving, it becomes very unstable and likely to fall over, mainly because of its leg segment lengths in proportion to the overall structure, and the absence of balancing mechanism, the structure fails to lift one leg but keep the rest of the body stable, other problems including inaccuracy in production and frictions between components, and the increasing weight of the structure due to more rigid elements added. The purely mechanical puppetry approach to achieve walking movements appears very difficult to succeed in this case, and due to the bespoke 3D-printed joints used for producing each prototype, the speed of physical iterations is very slow, an effective measure to quickly explore and test other possible solutions should be developed. We adopted the idea from the Passive Dynamic Walker of using wide feet in an arc-shape into our prototype 3, in order to stabilise the structure and prevent it from falling over.
æˆ‘ä»¬æ”¾å¼ƒäº†ä¹‹å‰æ›´åŠ ä»¿ç”Ÿçš„åŽŸåž‹è®¾è®¡ï¼Œå°±åƒäººä½“ç»“æž„ä¸€æ ·ï¼Œè¡Œèµ°çš„åŠ¨ä½œéœ€è¦è®¸å¤šæ¡è‚Œè‚‰åŒæ—¶åä½œå®Œæˆï¼Œå¦‚æžœè¦æŽ§åˆ¶å¤§é‡çš„PAMï¼Œæ‰€éœ€æ±‚çš„æŽ§åˆ¶é˜€æ•°é‡å°±éžå¸¸å¤§ï¼Œè€Œæˆ‘ä»¬æ‹¥æœ‰çš„æŽ§åˆ¶é˜€æœ‰é™ï¼ŒåŒæ—¶ä¹Ÿä¸ºäº†è¾¾åˆ°ä¸€ä¸ªéžå¸¸ç®€æ´é«˜æ•ˆçš„kinetic structure, æ‰€ä»¥æˆ‘ä»¬è½¬å‘äº†ç®€å•å‡ ä½•å½¢ä½“ä¸Žæ¨¡å—åŒ–çš„è®¾è®¡ï¼ŒWe arrived at a final design of an autonomous walking robot. The main body of the robot is a solid cube, with 4 contactable legs located at 4 vertexes of the cube, 2 at the top opposing vertexes, 2 at the bottom opposing vertexes in mirrored positions with the 2 top vertexes. In this arrangement, when one of the bottom legs is stretched out, the structure is able to roll over in the opposite direction of the leg, in the mean time, the stretching of one of the top legs in the same direction of rolling will secure the structure to a steady landing.
In order to achieve autonomous locomotion, we will need to add a gravity sensor on the robot, as when the robot complete 1 movement of rolling over, the whole structure turns upside down, the robots needs to know the updated status of top and bottom, so the control program can decide which legs to actuate for the next walking movement based on the updated position.
Note that the sensor morphology alone does not tell us very much; it is only if we take the specific interaction with the environment into account (which includes the actions of the motor system as well) that we are able to understand the role of morphology in behavior.
Not only do the retinas of insects have non-homogeneous morphology, but the retinas of mammals, including humans, are heterogeneous as well: The spacing at the center is more dense than on the periphery, which is unlike standard cameras, in which the distribution of the light-sensitive cells is homogeneous. One of the reasons why animals can process visual signals so rapidly is that the retina already does a lot of preprocessing before the signals are sent on for further processing. This massively parallel peripheral processing ability is crucial to achieving real-time behavior. And in order to do this processing, morphology has to be exploited. This idea is also exploited in the design of artificial retinas and has a long history, as well as generally in the field of space-variant sensing.
This thesis has tried to discuss the results of many different kinds of studies in the field of kinetic structure and morphology, we have argued that the emergent potentials of morphology, material and their association with the structural composition can facilitate major impact on the control system of kinetic structures. There is a lot of research on materials, performance dynamics, adaptive behavior and control that we would want to include eventually into our considerations. From a design point of view, this will undoubtedly lead us to design different and hopefully better kinetic structures for specific tasks. The next stage of this our design thesis will focus on the design and observation of adaptive behavior through interaction with people and physical environment.
...(download the rest of the essay above)