We have developed autonomously moving pine-cone robots, which are made of multiple joined pine-cone scales for outdoor natural environments. We achieved these natural robots by using pine cones as both natural hygromorphic actuators and components of the mechanisms. When they are put in outdoor places where moist periods (e.g., rain) and dry periods repeatedly occur, they can move up and down on the spot or move forward. This article describes the motivation behind our research, the design and implementation of three different hygromorphic actuators, and applications for autonomously moving robots in nature.
There is growing interest in constructing artificial life in the real world. One of the merits of doing so would be that we can understand the essence of life better; for example, life is normally wild and uncontrollable . Here, research attempting to realize artificial life in the real world can be roughly divided into those using electronic and mechanical hardware [11, 19] and those using chemical materials [1, 23]. Moreover, some works of art have tried to realize artificial life in the real world intentionally or have done so unintentionally. Such work can be characterized by autonomous behavior that looks as if it had its own volition. What makes it different from research is the variety of approaches used to create it—not only hardware [12, 24] and chemicals [6, 7] that are often used in research, but also natural objects or natural energy [14, 16, 20, 21].
We view the characteristics of natural products that don't have the same shapes or stable natural energy supplies as naturally leading to autonomy involving wild or uncontrollable behavior. Most work of this kind either has used natural objects as materials or has driven artificial objects with natural energy. However, by combining these two methods, we aim to achieve a novel form of artificial life derived from fully natural products. This study is also an attempt to pose a new question: whether a sense of life arises when a reconstructed natural object moves autonomously in a different way from its original form and behavior.
In order to achieve a novel form of artificial life derived from fully natural products, we initially focused on natural hygromorphs, that is, objects that change their shapes when the environment changes from humid to dry (and vice versa). We can use such natural objects as both natural hygromorphic actuators and components of the mechanisms. Several plants, such as wheat awn and orchid tree seedpod, have remarkable hygromorphic functions . Among them, we focused on pine cones and their hygromorphic function because pine cones are strong enough to withstand various movements. Moreover, it is known that an individual scale separated from the pine cone's shaft transforms in the presence of moisture. We used scales to develop three types of hygromorphic pine cone actuators: a basic opening/closing actuator, a height-change actuator, and a moving actuator. We did so by designing the joint pattern and interconnection systems of the scales. This article describes the transformation characteristics of the scales relative to the moisture content and time, as well as the design of three hygromorphic pine-cone actuators and applications for autonomously moving pine-cone robots in nature.
2 Related Work
2.1 Artificial Life Consisting of Natural Materials and/or Moving Autonomously with Natural Energy
The pioneering work realizing a kind of artificial life that moves autonomously with natural energy is Theo Jansen's strandbeest , which is composed of polyvinyl chloride pipes and which can move smoothly like an animal using wind power. Rolling Shadows , which consists of hundreds of miniaturized solar toy cars filling the shadows of pedestrians who are standing still, moves autonomously and generates chaos when the pedestrians walk away and the sunlight hits the solar cells on the cars. Additionally, as an example of artificial life that consists of natural materials, Yamaoka's Walking Tree , which uses an electronic actuator for power, incorporates various natural materials (such as branches) as feet for the robot. Another such example is Maekawa's Walk , which also uses electronic actuators and natural branches as components of the robot's body; it moves autonomously by creating a model of a robot based on 3D scanning of the branch on the spot and conducting reinforcement learning. In contrast to these previous studies, the authors propose a pine-cone robot that uses pine cones both as natural hygromorphic actuators and as components constituting its mechanisms.
2.2 Mechanisms Inspired by Biological Hygromorphs
There are various artificial mechanisms inspired by the transformability of biological hygromorphs, including the pine cone. For example, there is a smart fabric that automatically opens itself when it gets wet from sweat from the person wearing it and closes when it dries . The water-reactive architectural surface  is a laminate water-reactive material that automatically detects humidity and changes its shape without mechanical or electrical elements. Shin's hygrobot , which is composed of a hygroscopically responsive film, is a self-locomotive ratcheted actuator that uses environmental humidity. The mechanisms described so far use artificial materials to realize the functions of hygromorphs. On the other hand, Holstove et al. utilized natural materials, such as special cut veneer, both as natural hygromorphic actuators and as components . Our pine-cone robot, which utilizes pine cones both as natural hygromorphic actuators and as components of its mechanisms, differs from previous work of this type in that it can autonomously move forward in addition to moving on the spot.
3 The Principle of the Pine-Cone Hygromorph
Figure 1 shows the principle of the pine-cone hygromorph. The pine cone changes shape depending on whether it is absorbing moisture or drying. As shown in Figure 1(a), each scale is composed of different fibers on the inside and outside. Because the outer fiber is a thick water-absorbing vascular bundle, when dried, it shrinks greatly, causing a pulling force that opens the scales. In contrast, when the outer fiber gets moist, the pulling force eases and the scales close. Moreover, as shown in Figure 1(b), an individual scale separated from the pine cone's shaft can transform its shape with moisture.
4 Experiments to Reveal the Characteristics of Pine-Cone Hygromorphs
We performed three experiments to reveal the characteristics of the pine-cone hygromorph: one to measure the transformation angle as a function of moisture content, another to measure the transformation angle as a function of time, and a third to measure the transformation angle when two scales are artificially joined. Whereas the pioneering experiment of Reyssat used Pinus coulteri , our experiments characterized the properties of different kinds of Pinus thunbergii, which inhabit the authors' country (Japan). Additionally, we revealed new properties such as the transformation angle of two artificially joined scales. Here, we denote by θ the angle between the state when the scales of pine cones are dry and when they are moist as in Figure 1(b).
4.1 Experiment 1: Transformation Angle as a Function of Moisture Content
This experiment measured the transformation angle of a scale separated from the pine cone's shaft as a function of its moisture content. We set five moisture-content conditions ranging from 100 to 500 mg in 100-mg steps. Moisture was supplied to the outside of the scale by using a dropper. The transformation angle for each moisture content condition was measured. This measurement was performed three times for three scales. Each trial was conducted over 6 h or more for each scale, until the scale, which had absorbed water, was sufficiently dry.
Figure 2(a) shows the average transformation angle and the standard error in their dependence on moisture content. The scale transformed its shape with even a small amount of water, such as 100 mg. The transformation angle saturated at a moisture content of 300 or 400 mg or more, and the angle at that time was from about 40 to 50 deg. Although the angle was slightly different on each scale, these results suggest that the transformation angle becomes larger when the amount of water supplied is increased, and it saturates at a certain amount.
4.2 Experiment 2: Transformation Angle as a Function of Time
This experiment measured the transformation angle of a scale in its dependence on the elapsed time after a certain amount of water was applied to it. On the basis of the results of Experiment 1, the moisture content was from 300 to 400 mg, values at which each scale had reached the maximum θ in the first experiment. This measurement was performed three times on the same three scales used in Experiment 1.
Figure 2(b) shows the average transformation angle and the standard error in their dependence on the elapsed time. The plot shows that it took from 30 to 70 min to absorb the moisture fully, and more than 220 min for the scales to dry. These results show that the scale takes much longer to dry than it does to absorb water.
4.3 Experiment 3: Transformation Angle When Two Scales are Artificially Joined
This experiment measured the transformation angle when two scales were artificially joined and sufficient moisture given. Several artificial joining methods were tried, but eventually we decided to use a glue gun for ease of adhesion and because it did not scratch the scale. We prepared four sets of two scales joined in the open state and supplied them with sufficient moisture. The average transformation angle obtained in the experiment was 95 deg, which was within the range obtained by doubling the range of the maximum transformation angle of one scale. These results suggest that the influence of the joining is small.
5 Pine-Cone Actuators
We devised three types of pine-cone actuators: a basic opening-closing actuator, a height-change actuator, and a moving actuator. Each actuator utilized the pine cone's hygromorphic function. We designed the joint pattern and the interconnection systems of the scales for each.
5.1 The Basic Opening-Closing Actuator
The opening-closing actuator is opened and closed reversibly by supplying it with water and drying it naturally. Two types of actuator can be considered, according to the direction in which the scales are joined. The first type has two scales open, and the actuator in this case closes after absorbing moisture, the same as the one we tried in Section 4.3. The second type has the outsides of the scales back to back; it opens when it absorbs moisture.
On the basis of the aforementioned design guidelines, two types of opening-closing actuator were realized, in which two scales were joined with glue. Figure 3 shows how the basic opening-closing actuator operates. We confirmed that the behavior during moisture absorption and drying reverses depending on the type of actuator, as shown in Figure 3(a) and (b). For the first actuator, the average angle between the insides of the scales for four pairs supplied with sufficient moisture was 85 deg. For the second actuator, the average angle between the outsides of scales was 91 deg.
5.2 The Height-Change Actuator
The height-change actuator changes height repeatedly by supplying water to it and naturally drying it. In order to make the scale stand stably on the floor, it is necessary to join three or more scales with glue and place them so that their insides are in contact with the floor. Additionally, by joining together more scales to lengthen each leg, the height at which the actuator can stand can be increased.
Accordingly, we made height-change actuators with one to three segments. Figure 4 shows how this actuator operates. Figure 4(a) shows the dry status, and Figure 4(b) shows the wet status. Table 1 shows the height change measured twice and averaged over three actuators with one-segment legs (three pieces), two-segment legs (six pieces), and three-segment legs (nine pieces). The change in height varied by a factor of about two or three, depending on the number of segments.
|No. of segments .||No. of pieces .||Height (mm) .|
|No. of segments .||No. of pieces .||Height (mm) .|
5.3 The Moving Actuator
The moving actuator autonomously moves as long as moisture is supplied to it and it dries afterwards. We supposed that the moving actuator could be achieved by (1) joining two or more scales in a bilaterally asymmetric shape, (2) turning off the hygromorphic function of the right scale, (3) placing these scales on a sawtoothed floor, and (4) supplying moisture to the scales and drying them in air. When moisture is supplied, the actuator lifts its feet so as to slide on a gently sloping part of the floor surface; when the actuator is dried, it pulls down on a steeply sloping part and pushes itself forward.
Based on the aforementioned design guidelines, we implemented prototypes consisting of a sawtoothed floor and joined scales. A sawtoothed floor surface with a 4-mm interval was made with a 3D printer, and it was used for testing all prototypes. For the joined scales, three prototypes were made, which differed in the way the scales were joined in a bilaterally asymmetric shape and the way the hygromorphic function was turned off in the right-hand scale. In the first prototype, shown in Figure 5(a), scales were joined in a shape like a letter Y, and the scales on the right side were waterproofed so that they lacked the hygromorphic function. In the second prototype, shown in Figure 5(b), the scales were also joined in a shape like a letter Y, and the right side had scales made from an artificial material. The artificial material was PCL, a biodegradable plastic, so that the pine-cone actuator would eventually return to the environment after repeated autonomous movements. In the third prototype, shown in Figure 5(c), the right side had a triangular shape with a slightly bent tip and was made from PCL. We confirmed that all of the prototype systems could move autonomously when moisture was supplied and while they dried. For prototypes 1 and 2, the forward movement occurred once or twice in a row before falling over. For prototype 3, the forward movement continued stably.
Therefore, we examined in detail how the actuator would move over time for prototype 3. First, we recorded and analyzed a real-time video of the first forward movement  and a time-lapse video of three forward movements . We then found that the forward movement was divided into four steps, and the displacements differed slightly between the first step and subsequent ones. Figure 6(a) shows the first forward movement in four steps: ① the system opened fully in a totally dry state, ② it closed after absorbing moisture, ③ it opened slightly after drying, and the natural scale (left leg in Figure 6) sat on the sawtoothed floor, and ④ the artificial scale (right leg in Figure 6) moved slightly to the right. Figure 6(b) shows a standard movement after the first forward movement, which was different in that it did not start from a completely open state. Second, we measured the distance moved over time after moisture was supplied. Here we define the displacement d as the amount by which the natural scale moved from step ① in the first forward movement. Similarly, we define the displacement d′ as the amount by which the natural scale moved from step ①′ in the standard movement after the first forward movement. This measurement was performed three times each for d and d′. Figure 7 shows the average displacements and the standard errors as functions of time. The plot shows that the averages of d and d′ stabilized at slightly smaller distances after reaching the maximum. The values of d and d′ at this time were about 15 mm and 9 mm respectively, as shown around the 50-min point of the graph. This means that d tends to be slightly greater than d′. Analyzing the graph results in terms of the four steps mentioned above, we find that the averages of d and d′ were at a maximum in steps ② and ②′, respectively, and then stabilized at slightly smaller distances in ③ and ③′, and so on.
5.4 Experimental Results and Discussion
We designed and implemented three types of hygromorphic actuators made of multiple joined pine-cone scales. We confirmed that each actuator worked, although under limited conditions. With regard to the open-close actuator and the height-change actuator, we examined only the case where sufficient water was supplied and the transformation was performed at the maximum angle. However, considering the experimental results described in Section 4, we consider that it may be possible to control other transforming angles and height changes by changing the amount of water given. With regard to the moving actuator, the number of continuous movements depended on the prototype. One of the reasons for this may be that in the case of the Y-shape (prototype 1 and prototype 2), the scale on the right side did not fit the uneven intervals of the surface and it was easy for it to lose its balance. In this implementation, a stable forward movement was realized by modifying the shape on the right side as in prototype 3, but prototypes 1 and 2 may also work well if the spacing of the unevenness of the sawtoothed floor surface is changed. With regard to prototype 3, we evaluated the average movement distance over the elapsed time for one forward movement. We also confirmed visually that this forward movement was stable and continuous. In the future, it will be necessary to quantitatively evaluate the upper limit of the number of continuous movements and the cumulative movement distance by increasing the length of the sawtoothed floor surface.
6 Possibility of Introducing a Pine-Cone Robot as Artificial Life in the Real World
Here, we discuss the possibility of using three actuators in artificial life in the real world.
6.1 Application Examples
Two kinds of application design for autonomously moving a pine-cone robot can be considered, depending on the water supply method. The first is an application in which the autonomously moving pine-cone robot is in a place where water is regularly obtained, and the second is an application in which the pine-cone robot “lives” autonomously outdoors.
As an example of the first application, we placed the height-change actuator in a flower bed and confirmed that it could move up and down on the spot by absorbing water supplied to the flowers, as shown in Figure 8. If we choose plants and seasons that match the time of the pine-cone robot's transformation and the time it takes for watering them, the movements of the robots may serve as a timer for watering the flowers. This application would be a kind of symbiosis of natural and artificial life. For the second application, we are considering placing the moving actuator outdoors on uneven surfaces and in places where the humidity changes significantly (for example, due to rain or morning dew). As shown in Figure 9, we have confirmed that the pine-cone robot can move on an artificial surface with random irregularities indoors. In the future, we will investigate the possibility of it moving on natural surfaces with random irregularities such as stones and gravel and with water obtained from the environment.
6.2 Autonomously Moving Pine-Cone Robots as Artificial Life in the Real World
To determine whether people perceive an autonomously moving pine-cone robot to be lifelike, we collected opinions by giving poster presentations and showing a video at domestic and international conferences [17, 18]. We made a time-lapse video of the pine-cone actuators .
Those who understood the nature of the pine-cone robot after viewing the poster and the video seemed to be surprised by the fact that it could move autonomously. Many of those who attended the presentation expressed opinions like “the autonomous motion is funny and cute,” “I like this autonomously moving pine-cone robot,” and “it looks something like a creature.” These opinions indirectly or directly indicate that people feel there is something lifelike in the autonomously moving pine-cone robot.
Our pine-cone robots have some limitations. First, the proposed application in which the robots “live” autonomously outdoors has not been fully realized. We will investigate the possibility of the robot getting natural moisture from the environment and moving on natural surfaces with irregularities such as stones and gravel. We expect there will be many problems in outdoor environments, such as the lack of uneven surfaces of a suitable scale and the presence of too much or too little moisture. We will conduct mid- and long-term experiments outdoors to examine these possibilities and determine whether they should be regarded as positive factors causing uncontrollable movements in the pine-cone robot or negative factors to be avoided.
Second, although pine-cone robots are supposed to eventually return to the environment after repeated autonomous movements, not all of their materials are necessarily biodegradable. In particular, we should investigate biodegradable replacements for the bonds for the scale materials and the waterproof and water-repellent materials used in the prototypes.
Finally, we only tested how people perceived the autonomously moving pine-cone robot by showing them a video at a poster presentation. In the time-lapse video, the pine-cone robot seemed to move faster than it actually did. It will be necessary to determine whether people actually perceive the robot to be lifelike even when it is moving at a slow pace.
8 Concluding Remarks
We implemented three types of hygromorphic actuators made of multiple joined pine-cone scales. Moreover, we showed the possibility of deploying autonomously moving pine-cone robots incorporating these actuators in nature, where moist and dry periods repeatedly occur, as forms of symbiotic and autonomous applications. Our future work will be on building a moving actuator that can move forward autonomously on natural ground surfaces with a natural moisture supply, constructing a pine-cone robot made completely of biodegradable material, and determining whether people perceive the pine-cone robot to be a lifelike thing in motion.
This article is a part of the outcome of research performed under a Waseda University Grant for Special Research Projects (project number: 2019C-202).