Plants are sessile, autotrophic, eukaryotic organisms that are crucial components of almost every ecosystem and provide multiple resources to humans. Since the Neolithic era, humans have struggled to improve plants, to adapt them to better address their needs. Recently, farming technologies and breeding practices, which include molecular biology and genomics, have been greatly developed (Kanchiswamy et al., 2015; Poland and Rutkoski, 2016). Robots and advanced biosensors are already to use in the agriculture. The development of biohybrids, which are composed of biological and robotic parts, could become a smooth continuation and the next step (Gund, 2015; Ledford, 2017). There is also a growing understanding of biological life that pushes toward utilization of the advantages of combining living organisms and robots within one “being.” Is it feasible to increase the survival rate of cultures, crop yield, and even extend plants “functionalities” by creating biohybrids? The motivation coming from the rising food demand, the spread of great metropolis, and environmental challenges, began the struggle for the creation of robot–plant biohybrids, which will benefit from the synergy between both biological and artificial parts. For crops, biohybrids will help with biotic and abiotic stress management: robotic components could provide the plants with additional sensing, communication, or increased survival capabilities. Moreover, robot–plant biohybrids might become valuable elements of our urban space, even as a promise of living buildings, or at least as an inspiration for using continuous building material (Heinrich et al., 2016, unpublished manuscript¹).

Robotic prostheses are examples of biohybrids where feedback communication loops take seconds. After that, human motor or sensual capabilities are restored, the feedback is visible, and corrections can be implemented (Halloy et al., 2013; Bensmaia, 2015). It is worth to underline that many other investigated biohybrids systems exploit nanotechnology and soft robotics as tools, which allow them to efficiently interact with a living cell or an organism (Sicard et al., 2010; Ricotti and Menciassi, 2012; Patino et al., 2016). In robot–plant biohybrids, the most intriguing is the lack of nervous system and motor skills, although only a few fast motor plant reactions resembling animals motions are known. The motions of mimosa (Mimosa pudica), Venus flytrap (Dionaea muscipula), pea’s (Pisum sativum) tendrils are good examples of that and have been noticed by engineering due to their interesting mechanism of enwinding (Gerbode et al., 2012; Guo et al., 2015). However, despite being driven by different mechanisms than animals, behavior, decision-making, and even learning, are phenomena that exist in plants (Cvrčková et al., 2016; Gagliano et al., 2016; Schmid, 2016). It has been reported that cognition is restricted not only to brain-having organisms and that some of the complex molecular interactions seem to give a sufficient foundation for non-neuronal cognitive processes (Calvo and Baluška, 2015; Baluška and Levin, 2016). An escape from an illuminated area is an example of behavior that is similar, even on the level of molecular mechanism, in an animal Caenorhabditis elegans, with its primitive nervous system, and in roots of Arabidopsis thaliana, which lack nervous system (Yokawa and Baluška, 2015). The more general overview, where biorobotics and plant associations have been extensively discussed, had led to formulate the concept of biomechatronic systems and inspired the work on a plantoid robot (Mazzolai et al., 2010, 2014). It was proposed that robots with bioinspired structures or functionalities could be used for testing the biological hypotheses in a more reliable way than using the in silico experiments (Mazzolai et al., 2010; Schmickl, 2011; Zahadat et al., 2015). Bioinspired robots with different root-like functionalities have been under development, but until now, a few versions have been presented (Sadeghi et al., 2014, 2016). The non-random root behavior has led to the development of preliminary plant computing system that are based on root logical gates, which are a new development in natural computing research area (Adamatzky et al., 2016). Swarming was proposed as an experiment-based model for coordinated growth of roots (Ciszak et al., 2012; Kawano et al., 2012), and similar models could have an influence on the functioning of future biohybrids controllers. In this article, we focus on possibilities of communication between robots and plants, which emerge from biological processes existing in planta. We indicate processes on the level of interorganismal communication (between plants and other plants, animals, or microorganisms) that could be used for the robot–plant biohybrids development. We discuss also different models representing plant development, responses to environment, and structures, which could become useful for the development of controllers for biohybrids. Moreover, we present some of the already available devices, mostly agricultural sensors, which will be useful for robot–plant biohybrids development.

Tropisms are plant responses to external stimuli, which depend on stimuli’s direction. The tropisms improve plant adaptability by fitting plant growth to environmental constraints and localized resources. Tropisms could become beneficial also for robot–plant biohybrids. One of the proposed biohybrids tasks would be to create growing architectural artifacts, where biological symbionts, as well as robotic partners, collaborate in the process of building a structure (Hamann et al., 2015; Wahby et al., 2015). For this reason, biological processes that enable automated shaping of plants could become crucial for biohybrids functioning.

For many years, the idea of “talking trees” was in contradiction with human common sense. Plants were described as simple living automats, which in the best case could only gather and use information from the environment but could not communicate with other organisms. However, plants are capable of both sensing their environment and actively changing it (Kegge and Pierik, 2010). The first study on plant communication was reported in early 1980 (Baldwin and Schultz, 1983; Rhoades, 1983). A higher level of herbivore resistance was described in conspecific plants growing in close range to herbivore-attacked ones than in plants growing further away. Since then, the concept of plants interacting with each other, as well as with microorganisms and animals, have become well established (Kesselmeier and Staudt, 1999; Dicke and Baldwin, 2010; Holopainen and Blande, 2012; Karban et al., 2014; Mescher and Pearse, 2016; Schöner et al., 2016).

We suggest innate plant communication mechanisms with other organisms and environment as the most convenient approach to establishing an artificial connection between plants and robots. Human ability to transform plant signaling pathways is far more limited than the modification of robots. Hence, adjusting robots for perceiving plants derived signals and transmitting to plants nature-like compounds should become a priority in establishing plant–robot communication.

It has been discussed for years, how to utilize natural plant communication in agriculture, forestry, and architecture. One of the obstacles is a complexity of signal signatures and diversity of possible responses. Since the first studies of the role of volatiles in survival and adaptation, BVOCs has been successfully used in pest insects control (Sobhy et al., 2014) and in establishing crop resistance to abiotic stresses. Possible application of natural plant signals as pest control mechanism have become an interesting solution to the problems related with growth of human population, increasing food demand and well-known toxicity of the synthetic pesticides. At this moment, there is a growing number of “push and pull” approach initial tests (Du et al., 2016; Zhou et al., 2016); however, only one successful implementation was reported at large scale (Cook et al., 2007). A successful method based on the integration of stimuli that act (e.g., BVOCs) to make protected crop unattractive to the pest, while luring pests toward a different target that was chosen by the farmer (Cook et al., 2007). While most of “push and pull” tests utilize natural plant abilities, a first field test using genetically modified plants emitting insect pheromones was also performed (Bruce et al., 2015). Although, approaches to using GMOs in pest control could be more efficient, such as luring insectivorous or parasitoids by plants with an enhanced level of BVOCs emission (Kappers et al., 2005; Kos et al., 2013). Artificial emission of specific volatiles has been also investigated, what suggests possible future utilization of BVOCs in biohybrids (Kelly et al., 2014). It was shown that robotic mechanical wounding could stimulate plant defense mechanisms, like BVOCs emission, although, in a different way than an insect attack. Two applied types of mechanical damages resulted in activation of different types of signaling events (Bricchi et al., 2010). It proves the complex mechanism of regulation of BVOCs emission and suggests that use of artificial tools to create a certain plant response has to be more thoroughly investigated.

Nevertheless, in the perspective of biohybrid-plant communication, these are preliminary, but promising results and actions, which focus on BVOCs in plant communication happening aboveground. Detection of all compounds from BVOCs mix emitted by a single plant is still a very sophisticated process (Materić et al., 2015). Gas chromatography (GC)-mass spectroscopy (GC-MS) is a gold standard in chemical analysis of BVOCs, even very similar compounds can be separated. Even though downsizing is possible (Tridion-9 GC-MS), measurements are far from real-time detection (Materić et al., 2015; Niederbacher et al., 2015). The technique allowing BVOCs measurement in real-time is ion flow tube-mass spectrometry (SIFT-MS), which is direct and quantitative and does not require calibration against gas standards. On the other hand, isoprene and monoterpenes cannot be measured simultaneously, and different isomers cannot be distinguished (Materić et al., 2015). Another possible technique is proton transfer mass spectrometry (PTR-MS), the most sensitive real-time technique, but identification is limited to the nominal mass (Schaub et al., 2010). Although, some modifications of this method or coupling with GC can discriminate single compounds (Misztal et al., 2012). The even more challenging task is the measurement and detection of plant signals underground since the soil is far much more dense and heterogeneous than air, and humidity changes can affect the diffusion and distribution of infochemicals. Recent approach to cope with these impediments is inserting sampling devices in the soil, next to the root, but most of them cannot provide real-time measurement and cannot detect signals transmitted through CMNs (Blande and Glinwood, 2016). Another example of detection technique is bi-enzyme biosensor using carbon nanotubes (CNTs), which is able to detect in nanomolar range methyl salicylate (MeSA), one of the plants common BVOC. This bi-enzyme biosensor consists of electrochemical biosensing platform with immobilized enzymes (salicylate hydroxylase and tyrosinase), as recognition elements. Enzymes are immobilized onto CNTs matrix on dedicated carbon electrode surface. This biosensor selectively detects MeSA, which is known as a potent VOC released by plants mainly under biotic stress. This technique could be applied as a swift method for detection of pathogen infections before harsh phenotypic changes had occurred (Fang et al., 2016a,b). E-noses were also shown to detect and discriminate BVOCs, so multiple technologies could be investigated for being applied as a tool for detection of BVOCs derived from plants in biohybrids (Ghaffari et al., 2012). Such capability of robot–plant biohybrid to detect early infections will help in keeping plant fitness on both, organismal and ecosystem levels.

Although it could seem discouraging in creating plant–robot communication, we do not have to measure and differentiate all BVOCs and other infochemicals simultaneously. We should focus on examined, most important and influential compounds, creating suitable, miniaturized, and real-time sensors. This approach, as well as crucial volatile compounds emission devices could allow to establish a two-way communication in plant–robot biohybrids. The goal is to use robots that could not only detect changes in plants but ones that could also modify plant metabolism, development, and defense response through mimicking and amplifying natural plant signals. Bearing in mind that robots could communicate and coordinate actions on a relatively large area, it could provide a great tool to control not only single plant, but to govern a whole field or ecosystem at the same time.

Plants, although immobile and not moving rapidly, can react quickly to changing environmental conditions. Information about some of the environmental changes that are perceived locally has to be transmitted to other distal organs to allow for an adaptation. Attack of a herbivore or a mechanical damage caused by the wind are examples of events that require a fast reaction on the local and whole-organism levels. Different types of rapid long-distance signaling pathways ensure an efficient communication between distal organs and make it possible to develop a complex system of “signatures,” which can lead to a proper response to a stress factor. Deciphering plants’ reactions that emerge rapidly after stimuli would offer a chance for quick adaptation of robotic elements of biohybrids to changes in the environment and to plant-counterpart’s physiological conditions.

Information flow from plants to robots could depend on diverse plant processes, some of which could be monitored by wet-lab analysis or use of electronic sensors. Laboratory-like actions are not easily applicable in biohybrid conditions. Sensors that would be able to measure plant’s relevant physiological variables may be used for collecting data from plants in biohybrids. This becomes very useful for establishing a communication channel between artificial and natural elements. It is important that only non-destructive methods of measurement of physiological processes should be considered; invasive, destructive methods could influence plants in many undesirable ways (Busemeyer et al., 2013; Tattaris et al., 2016). There are many sensors available on market, which enable to measure different agriculturally relevant plant features. Decagon Devices (now Metergroup) company is specialized in sensors, which are designed to collect data from the forest canopy. Ready sensors allow, for example, to measure leaf area index (LAI) (Sone et al., 2009), stomatal conductance measurements, electromagnetic radiation reflected from canopy surfaces, leaf wetness, tree or branch circumference, surface temperature, and more. There exist also a ready microclimate monitoring station and a set of sensors, which deliver information about the soil, and irrigation. Yara Water-Sensor monitor water status of the crop by measuring the turgor pressure in leaves, hence it helps in irrigation’s automatization and could be used even to detect hydraulic signals (Yara, 2017). Data collected from these sensors, together with climate and soil data may strongly facilitate designing, development, and implementation of electronic elements in robot–plant biohybrids. Such sensor system would enable the reliable observation of robot’s impact on plants, one of the key feature for a working biohybrid. Development of biosensor plants in order to show changes in physiological status of plants in a way that would allow for easier data collection, for example, luminescence or color change, could open new possibilities (Feng et al., 2015; Medford and Prasad, 2016). Adaptation of some already existing laboratory methods to requirements of biohybrids could also allow for progress in the future.

One of the tools that could be used for organ specific delivery of effective compounds to plants are organic electronic ion pumps (OEIPs), a polymer-based delivery system and organic electrochemical transistor (OECT) (Simon et al., 2009a; Nielsen et al., 2016). They can deliver with high precision and efficiency a range of biologically relevant substances. Ion transport was shown to be able to mimic, and to precisely affect Ca²⁺ signaling in a programmed way (Simon et al., 2009a,b). It was demonstrated that OEIPs can be used as programmable “machine-to-brain” (or “machine-to-cell”) communication platform. Recent studies have shown that it is possible to couple OEIP with stimulation unit—a sensing unit, which may work in parallel (Jonsson et al., 2016). There are many more variants and different approaches to signal transfer—organic thin film transistors (OTFTs), polymer electrodes, smart textiles, ion bipolar junction transistors, chemiresistors—which could be applied to develop biohybrids. For example, OTFTs may be used as sensors for: pH, K⁺ ions, cysteine, glucose, lactate, DNA, humidity, and saline (Liao et al., 2015). It was also shown that in fixed plants xylem, a transport tissue in plants may be used as a wire to conduct electronic signals delivered by OECT (Stavrinidou et al., 2015). Nowadays, one of prominent research trend is to fit these devices for applications in neuroscience (BiOprobe Project, FP7-PEOPLE, Project ID: 300106).

Programmable ligand detection system in the plant, based on periplasmic-binding proteins and ligand receptors, was used to develop highly sensitive and specific sensor system (Antunes et al., 2011). These precise receptors, linked with plant visual response system (e.g., bleaching of the plant in the case of detection of ligand or luminescence), could be used for simple and safe detection of many receptor ligands—chemical agents, volatile compounds, or even explosives (like TNT—2,4,6-trinitrotoluene). It was shown that it is possible to develop transgenic A. thaliana and Nicotiana tabacum (called “detector plants”), which are capable of detecting TNT by roots and shoots. The response of plants to detection of ligand (de-greening) was visible after 24–48 h (Antunes et al., 2011). Recent studies have shown that wild-type plants can be also considered as pre-biohybrids—by using nanoparticles. Nanobionic-plant system was created; it is based on spinach plant (Spinacia oleracea) and nanoparticles—single-walled carbon nanotubes (SWCNTs) (Wong et al., 2017). This system was able to detect nitroaromatics in real-time and subsequently report that via attenuation of near-infrared (nIR) fluorescence in leaves. Nitroaromatics are transported from roots to leaves and then are aggregated in leaf tissue. That process results in relative changes in nIR emission intensity—which could be observed in real-time by nIR fluorescent nanosensors—SWCNTs (Wong et al., 2017). Using SWCNTs opens a lot of possible ways in the design and implementation in robot–plant biohybrids. New functionalities could be added, which will help in obtaining plant’s physiological status or environmental traces, like the presence of TNT.

Design and development of devices called electronic nose (e-nose) could also be useful in creating of biohybrid’s devices. There are a few examples of ready e-nose devices on the market. One of them is zNose^®, which could analyze and identify traces of organic compounds and can be followed by GC to improve the accuracy of the analysis (Electronic Sensor technology Inc., 2017). E-nose developed by NASA (called JPL Electronic Nose—ENose, third generation) is used currently on the International Space Station to monitor the environment in laboratories and detect air contamination from spills and leaks. The e-nose device based on multichannel quartz crystal microbalances (QCM) was also demonstrated; it contains polystyrene molecularly imprinted polymers layers as a recognition material. This device precisely detects terpenes emitted from fresh and dried plants from Lamiaceae family, what was demonstrated for monitoring of herbs freshness (Iqbal et al., 2010). E-nose device might be upgraded with various modules, for example, vulnerable thermometer, which eliminates temperature impact on measurements (Lin et al., 2016).

The diversity of known technologies to measure various plants traits looks promising. Biohybrids composed of plants and robots will be equipped with reliable sensors, often previously tested in agriculture. Further development of technologies will be greatly appreciated: new sensors, alike synchronization of known sensors into one sensing system, will make it easier to formulate conclusions about plant’s physiological status. Different methods, for example, machine learning or regression analysis, could be then used to extract information about vegetative status (Doktor et al., 2014).

Mathematical modeling fits into robot–plant biohybrid subject as an excellent tool for relatively fast prediction and validation of stimulants that would allow plants to grow in the desired manner. Robots need controllers, which could partially depend on plant’s model to achieve plant related targets, like shape, adaptation, size. Plants display numerous features that can be mathematically characterized, such as a number of leaves, petals, branches, etc. (Battjes et al., 2016). It seems that significant mathematical correlations can be found by the study of development and plant response to environmental cues at both structural and molecular level (Coen et al., 2004; Chew et al., 2014). At the heart of every computational model is an algorithm—a formula that describes step by step operations that have to be performed in order to depict growth patterns or simulate plant response to different stimuli (Françon, 1997). The main challenge in creating such mathematical model of a plant is to find an appropriate description of the local states that are changing in time (due to, e.g., cell division) and global alterations in the structure of the plant (Prusinkiewicz and Runions, 2012). From a technical point of view, one of the most important breakthroughs in the field was the introduction of so-called L-systems, which is widely used method describing linear and branching structures, where each element is coded as a symbol and all branches are marked in the form of brackets (Janssen and Lindenmayer, 1987). To better describe the system, each element might be bound to other features like length, diameter, or concentration of hormone.

Description of growth and shape itself requires quantitative description expressed in numbers denoting distances, angles, areas, and volumes, while, for example, the arrangements of the elements on the axis can be described without specific measurements (Pabst and Gregorova, 2007). Therefore, one of the most important features that needs to be taken into consideration, when constructing a model, is topology and geometry.

Topology is the arrangement of the various elements in a given system that can be mathematically shown in relations between algorithms describing each element. Although this feature remains constant under different stimuli, while geometric parameters are changing, it is very important to use appropriate topology in the model because the placement of elements later determines exact direction of information flow (Kholodenko et al., 2012).

Model of the overall growth of the plant can be created by integrating all parameters describing local growth (for example, branches, leafs, roots) with a use of tensors expressing length, area, and volume (Jensen and Fozard, 2015). While modeling growth, it is important to consider two viewpoints: the Eulerian and Lagrangian. The first one specifies the region of growth with relation to the external fixed system, while the second one specifies the growth of material element referring to its final version (Prusinkiewicz and Runions, 2012).

Changes in the shape of a certain plant can be caused by the growth of cells, their movement, and death that can lead to the arrest in plant growth (Steeves and Sussex, 1989). Therefore, modeling cell structure and its division is a first step toward elucidating more complex models at the tissue or whole plant level, such as the arrangement of branches, root development, forms of leafs. After integrating all this information, it will be possible to make conclusions about general growth of a given plant in specific conditions.

Plant growth regulation is a multi-level mechanism connecting molecular, cellular, organ, tissue, whole plant, and environmental level processes. Therefore, it is extremely difficult to elucidate exact stimuli that would allow the robot to control plant growth while maintaining symbiosis with it (De Vos et al., 2012). As was mentioned before, the starting point for studying growth in plants are cell pattern, division, and expansion growth models. From that base, every other higher level computational representation could be drawn (Prusinkiewicz and Runions, 2012). This is the reason that reliable model that describe even the growth of a single cell are required.

Plant cells are characterized by the presence of the cell wall. Moreover, they are deprived of cilium or flagellum, hence immobile, so the modeling of expanding growth is connected only to the change in size and shape. In order to model the expansive growth of cells, the equations need to be linked to biophysical processes. One of the most important factor is the turgor of cells, which is connected with water uptake and pressure against plasma membrane. If water uptake is perturbed, a cell’s wall deformation occurs. Hence, cell growth depends a lot on water uptake and transpiration rate on the organismal level (Ortega, 1985, 2010). The latest model explaining expansive growth is LOS model, which is strictly connected with the loss of stability theory. The assumption is that the turgor pressure is decreasing over time to a certain value after the decline of cell wall’s stability (Wei and Lintilhac, 2003). Another recent model incorporates calcium ion movement and chemistry of the cell wall component and suggests that cell wall expansion depends on the rate of the delivery of the substrate for constructing cell wall and pressure level (Proseus and Boyer, 2006).

The process of cell division was discussed as early as in the XIX century. At first, it was suggested that dividing wall was placed at a specific angle along the cell, then it was believed that dividing wall is actually meeting side walls of each cell at right angle, and then it was postulated that the shortest partitioning wall between two cells create two equal cells as a result of division (Sahlin and Jönsson, 2010). Computational models were created based on those assumptions, what showed that all previously suggested rules are needed, but their individual explanatory power is not enough (Sahlin and Jönsson, 2010; Besson and Dumais, 2011).

The right placement and orientation of division walls are the most important for tissue development models. Cells in those models can be represented as a set of points and growth simulated by addition of new points. A division that can occur in that model depends on the size of the structure created after several rounds of new points addition (Korn, 1969). In more recent models, cells in tissues are depicted by series of points with the application of the rule that allows for a random choice of division walls when they have a similar length (Dupuy et al., 2010; Besson and Dumais, 2011). Tissues topology, instead geometry, also can be taken into consideration, due to cells divide with regard to time rather than shape or size, what affect how combinations of L-systems can be implemented (Kennaway et al., 2011; Robinson et al., 2011).

One of the important geometric features of the plant structure is phyllotaxis. The geometric patterns that plants display with their leaves, shoots, or roots are originated from the specific placement of primordia (Braybrook and Kuhlemeier, 2010). What is interesting, angles between following primordia often assume the value of the golden angle (137.5°) or Lucas angle (99.5°) (Prusinkiewicz and Runions, 2012). An important question to be answered is how the positioning of primordia is determined, and then, how the initiation of development of specific tissues occurs. In the first model of this process, it was assumed that already existing primordia may inhibit the development of other initial primordia. Therefore, in the model, authors surrounded each primordium with inhibition zone and new structures were only allowed to develop when there was space for them (Mitchison, 1977). In following models, the distances between primordia and inhibitory effects of other primordia were taken into consideration. New structures were created after those effects weakened with time (Douady and Couder, 1996; Smith et al., 2006). All of these models were effective in recreating of organizations of primordia, while maintaining the golden angle between them.

If we consider a plant in its environment as a system, then the phenotype can be described as functional equations multiplied by parameters of the model and external variables in the system (Viaud et al., 2017). Light is one of the most important factors controlling plant’s growth. First simulation models taking into account the environment and light appeared in 1990s (Kanamaru et al., 1992; Měch and Prusinkiewicz, 1996). In those papers, authors proposed the construction of two separate models (plant and environment) and then simulating interactions between them. Communication between plant and environment was set up in the way that the plant sends information about the position of every leaf cluster and its radius and environment reacts to this by incorporating this radius into the light flow. In order to refer to the amount of light in the system, authors calculated the specific quantity of light that reaches plant (Kanamaru et al., 1992). Moreover, the plant model contains a few additional parameters determining the length of branches, the radius of leafs, etc. The simulation starts with a single branch, which supports leaf with an apex. Competition for light is manifested in two ways: either there is reduced branching or arrest in apices development—nevertheless, there is always a reduced amount of branches. In these models, plant responded to the amount and direction of light, which simulate heliotropism, but another stimulus could also be considered (Medina-Ruíz et al., 2011; Viaud et al., 2017). Modeling plant’s response to stimuli is more and more feasible due to the accumulation of plant science’s knowledge, hence, quantitative prediction of when and how a plant responds to different stresses is also becoming more reliable.

Mathematical models allow for prediction of the effect of certain stimuli. Simulation-based models specifically enable researchers to observe the effects on multiple levels of system organization (from molecular to tissue development) in regard to time. After knowing how the cells divide and grow into tissues, it will be possible to model plant organs such as leaves, branches, and roots. This scientific, mathematical approach to plant sciences will make possible to develop biohybrids controllers that could response to plant-derived information in a meaningful and reliable way. On our way to constructing biohybrids, not only engineering should provide more advanced tools but also biology should provide more understanding of plants biology, in a way that could be implemented within applied sciences.

Except mentioned earlier attempts and works on subjects associated with plants biohybrids, there are a few EU funded projects that are dedicated for plants bioinspired or plants biohybrids tasks. Two already finished EU funded projects are good example of pioneering and interdisciplinary work on plants and robots. PLEASED project (FP7-ICT, no. 296582) based on idea of using plants as biosensors in non-laboratory conditions, where measurement of changes in plants electric activity were tried to be deciphered and equate to the external events. Estimators for light stimulus on the base placed in plants electrode signals were developed and analyzed (Chatterjee et al., 2014). PLANTOID project (FP7-ICT, no. 293431) was focused on abstracting roots properties that enable effective functioning, and then bioinspired robots were made that could be used for testing biological hypothesis (Mazzolai et al., 2010; Sadeghi et al., 2014, 2016). The PLANTOID’s idea of “plants as robots and robots as plants” is still being intensively investigated and next outcomes and applications are expected. Flora robotica (Horizon 2020, no. 640959) is a visionary project aiming to create a highly integrated, symbiotic system of plants and robots in which robots enhance and control the development of plants. The main task is to construct physical structures, with further applications in architecture, resulting from plant growth controlled by robot-delivered stimuli (Hamann et al., 2015). The flora robotica is focused on interactions of robots and plants in order to create highly integrated robot–plant symbiotic system. Construction of robot–plant biohybrids, in which robots could guide the direction of plants growth and plants would provide growing structure to robots, is the main of the project’s goals: “the natural plants provide growth structures and sensation capability, robots provide extended sensing and decision-making capabilities” (Hamann et al., 2015). The tropisms were proposed to be used for shaping the plants in the desired way. For example, control by applying LED light over the growing tips of beans was shown, i.e., the directing of plant’s growth to achieve desired spatial targets. The RGB LED light was provided with different settings in order to evolve controllers more efficient in directing bean’s tip into the desired position. These results give background for the further development of biohybrid controllers, which will enable to shape plants (Wahby et al., 2015). Physiological sensors and structural elements that could help in creating robot–plant biohybrids are also under development.

When including also natural computing and part of nanotechnology research, these are signs of growing scientific community that is interested in plants biohybrids related subjects. However, time will tell us which technologies and ideas occur to be the most fruitful and which will be continued—recent state of the art does not allow us to make great prediction.

Advanced electronics in agriculture, science-based ideas concerning plants’ sensibility, a search of environment-friendly technologies, are all facts supporting the growing interest in robot–plant biohybrids. Living buildings and advanced agriculture are two examples of application of such hybrids that meet global challenges. In robot–plant biohybrids, the artificial part will detect significant changes in both plant physiology and the environment, but it will also purposefully manipulate the plant’s inner processes. We discussed the biological background for establishing plant–robot communication in Figure 1. Conclusions from an analysis of plant biology might direct further efforts concerning biohybrids. Such hybrids would function probably more on the “ecosystem” level and will depend on a massive cooperation between multiple plants/robots. Difficulties arising from decentralized regulation of plant growth and development and the lack of plant nervous system urge us to suggest various possibilities for gathering plants’ data and for influencing physiological processes. However, plants dissimilarity from animals does not change the abstracted life features. Moreover, molecular processes occur very fast and we should try to efficiently interfere with them, if we want purposefully influence decision-making of a living entity (Brunk and Rothlisberger, 2015; Schmickl et al., 2016). Relying on plants macroscopic feedback, which can take days or weeks (like developmental/growth events), could be risky, due to time that enables the appearance of many unpredictable factors that irreversibly influence the organism. Hence, sensing of plants’ biological processes should rely on early and key signaling pathways, which might potentially be used in biohybrids. Similarly, known channels for plant-other organism communication, like volatile compounds, should be exploited. Technological advancements are the keys for the future application of current biological knowledge for establishing plants-robots societies. Sensors will ensure the important link between the world of living organisms and artificial electronic robots. Needs for setting robot’s controllers in biohybrids might enforce implementation of some of the existing plant’s biomechanical or developmental models. Although the creation of robot–plant biohybrids appears very challenging, even now, we could underline some of the plant features that will be crucial for the following steps. As an example, plants ability to communicate with other organisms gives a base for establishing a plant–robot communication channels. Moreover, future developments will not only provide new technologies, but coincident use of biological models, advanced sensors, and methods for manipulating plant’s growth and development might lead to new discoveries concerning plants biology. Despite many challenges to overcome, the robot–plant biohybrids probably are to be soon intensively developed and investigated.

TS contributed to writing, correcting, and editing the paper. RK, WK, SW, AS contributed to writing the paper. PW contributed to supervising and correcting.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer LB and handling editor declared their shared affiliation, and the handling editor states that the process nevertheless met the standards of a fair and objective review.