General Literature Review

From Control Systems Technology Group
Revision as of 14:25, 4 June 2018 by S162286 (talk | contribs)
Jump to navigation Jump to search

Literature Review

The literature review is divided in two branches: one general literature review concerning itself with robotics technology and current methods used for reforestation and an extended literature review. The latter was held to zoom in on the specific case of reforestation methods, their effectiveness and evaluation parameters, with the goal of obtaining clear cut criteria for assessing the need for a robot. And if it turns out such a need arises to obtain insights into what functionalities the robot ought to have to outperform the current methods. This extended literature review can be found in Extended Literature Review. General information about the project can be found in PRE2017 4 Groep6.

Available robotic technologies

Initially our artifact-to-be-designed was envisioned as a mobile (semi-)autonomous mobile robot which would cover terrain in need of reforestation, evaluating the ground using sensors to obtain parameters which give information about the fertility of the ground (e.g. humidity, acidity, bacteria presence, etc.) and a planting mechanism to plant the seeds if the environmental conditions are deemed favorable. Preferably the artifact-to-be-designed would be a modular robot, consisting of a basic chassis upon which modules could be placed to add or interchange functionalities such that it can also be used in other areas besides reforestation. Therefore the corresponding literature topics were investigated, the results of these researched items can be found below.


Modular robotics is a useful tool in the design of robots for in-field applications, as building a functional specialised robot from scratch is a time-consuming and cost-intensive process. If a modular design approach is taken, the costs of designing a robot could be severely reduced as one general robotic platform with some general functionalities would serve as the starting point, upon which modules can be placed to give the end-product the desired capabilities. A drawback of this modular design method, however, is that the design space will expand explosively due to the seemingly limitless possible configurations the robot could have (Farritor & Dubowsky, 2001) [1]. However, this design space can be brought to proportions by severely reducing it, by placing the constraints which arise from the task to be completed by the robot onto the possible configurations (Farritor & Dubowsky, 2001) [1]. By doing so any and all designs with but a singular deviation which would compromise the execution of the task are immediately discarded in the earlier stages of development.

Some examples of robots which implemented a modular design and with similar environmental working conditions as our to-be-designed seeding robot include the Small Robotic Farm Vehicle (Bawden et al., 2014) [2], the 4-wheel steering weed detection robot of Bak and Jakobsen (Back & Jakobsen, 2004) [3], the Amphibious Locomotion Robot of Li, Urbina, Zhang and Gomez (Li et al., 2017) [4], and the Reconfigurable Integrated Multi-Robot Exploration System (RIMRES) (Cordes et al., 2010) [5]. These robots have in common that they are mostly based on a singular platform, suspended by wheels for locomotion, upon which several modules (e.g. sensors, mechatronic arms, pay-loads, other deployable robots, etc.) can be placed to increase functionality.

A special class of modular robots are the so-called self-reconfigurable modular robots which can change their shape to comply with dynamic environmental constraints and task requirements. Some examples of these self-reconfigurable robots include the I(CES) cubes (Unsal, Kiliccote and Khosla, 1999) [6], M-TRAN (Murata et al., 2002) [7], ATRON (Jorgensen, Ostergaard & Lund, 2004) [8], Modular Robot for Exploration and Discovery (ModRED) (Baca et al., 2014) [9], Polybot (Yim et al., 2003) [10]. Albeit this is an interesting topic of research, for our problem at hand it will not be a feasible solution, since most of these systems are on a mesoscale application, whereas the to-be-designed deforestation robot will be a macroscale prototype.

(Semi)-Autonomous Cars

A good description of the working of remote control systems is given by the patent on remote control systems, which is granted to Mitsubishi Electric Corp. by the US government (Hashimoto et al., 1996) [11]. This patent lists all the essential components for a remote-control system such as movement detector sensor, a transmitter and receiver unit, a display device to function as the user interface, etc.

Elon Musk, CEO of amongst others SpaceX and Tesla, leader in electrical and (semi-)autonomous vehicles, describes in his vision of the autonomous car in 2016, where software updates will dominate the improvement in functionality and degree in autonomy, whereas repairs by an actual mechanic will severely reduce. There is even the potential for turning non autonomous cars into autonomous cars by means of a software update (if the non-autonomous car has software capabilities). However, there may be some legal challenges involved in this method (Kessler, 2015) [12]. Another option for mobility, in the case of failure to implement a fully autonomous vehicle, would be remote control. An operational remote control robot is actually more closely related to a remote controlled toy car than an actual remote control car due to its size. The active patent for this is owned by Matsuhiro and shows the state of the art for these machines, which do not differ much from the state of the art for autonomous cars, considering a transmitter and receiver unit is the main component (Matsuhiro, 1984) [13].

An important aspect in autonomous vehicle is the path planning to get from A to B. An ambitious project, albeit one with high potential is to set up a communication network between other (autonomous) vehicles to share information regarding traffic densities, traffic jams and unforeseen obstacles due to accidents to get additional information for optimal path planning (Bagloee et al., 2016) [14].

Sensors for prospecting/evaluating ground

In order for to robot to be effective it should not waste time, energy and seeds by planting them in infertile soil, as after all the fertility of the soil will be the dominant factor determining the survival rate of the seeds and hence determines the effectiveness of the employment of our robot. Therefore, our robot-to-be-designed would severely benefit from sensors which can prospect the soil to some degree, at least to get a sufficient reading of the parameters determining fertility to rationally decide whether or not to commence the planting operation at a certain location.

Current technology allows the conditions of the soil to be inferred by a multitude of sensors. The simplest of them being a sensor which monitors a plant and determines whether the plant shows sufficient growth (Barnes et al., 2003) [15]. A lot of information can be obtained from the plant, like the salinity, nutrients and available soil moisture. However the application of this type of sensor would be inefficient and not desirable for our robot for two reasons:

  1. It takes a lot of time to monitor a plant and get sufficient data on its growth to be statistically significant.
  2. Our robot will be used for reforestation after forest fire, therefore most vegetation will have been destroyed by the wild fire, thus rendering next to no plants to take measurements of.

An alternative would be to use moisture sensors to determine the amount of water in the ground, since water is a critical component for a plant to grow, however water needs not be present in abundance as that could also be detrimental for plants (Kuang, 2012) [16]. Further sensors include NIR reflectance sensors. These sensors can accurately measure the organic matter within the soil. This leads to an accurate picture whether the soil is fertile enough to plant seeds, as the presence of bio-mass and micro-organisms indicate the existence of an micro ecosystem in which the plant can exchange nutrients for its growth. Vis-NIR sensors can also determine the amount of nitrogen and moisture in the soil. Which leads to an even more complete picture of the soil, as nitrogen is an important nutrient for plants.

Humidity in the air can also help determine whether the area is suitable. An RH sensor based on a Bragg grating can determine the relative humidity accurately. The optical fiber used to determine this can also house temperature, pH, pressure and more sensors (Correi et al, 2012)[17]. This results in a quite complete picture of the environment above the soil and can help determine the suitability for planting the seeds. The temperature sensor seems redundant as the average seasonal temperature will in general be known for a certain geographical area, and replanted trees will be growing in the same area as they used to grow thus we know that the climate is suitable, however it may give some important information about any and all remnants of the forest fire. If a small fire source remains which may not be visible with aerial/thermal imaging the robot can detects these fire seats and drive clear from them to prevent damage to the robot. Additionally it will be wise not to plant seeds near a still burning fire as the chances for survival will be low as it is most likely the seed will burn. A final benefit of employing a temperature is related to tertiary users; the robot could send an alarm to the fire department if a remnant of the fire is detected so that they can extinguish it before it can spread and turn into a second wild fire.

The robot can also be used in areas where the fertility is more or less predetermined, as some factors allow for an estimation of high or low fertility. Forest fires, for example, increase the nitrogen in the soil and in most cases the amount of carbon is also increased (Zabala, Celis Garcia and Lopez, 2014)[18]. This results in a soil that is suitable and fertile enough for the deployment of our robot without the need of inquiring information through dedicated sensors, which could potentially save a lot of processing time and hence make the operation of the robot more effecient.

Drilling/plowing/seeding mechanism

There are a lot of variants to keep in mind about the seeding mechanisms. Seedlings are often used as to plant trees, but is not viable to do on big scales. Due to both the fact that seeds are easier to transport, handle and plant than seedlings this should make up for the loss on the success rates (Atondo-Bueno et al.,2016) [19]. Setting and or measuring whether the ground variables like vegetation cover and soil moisture levels are within acceptable ranges will also greatly benefit the results. These variables vary a lot between species so they will have to be researched which are the best option for every species that will be planted in the new region. The article covers the effects of seed hydration, soil moisture levels and vegetation cover on the oreomunnea mexicana. For the respective species vegetation cover affected the emergence of seeds negatively, but soil moisture content affected it positively. Hydration did not make a significant difference. While the general effects will be the same for most species, optimal levels of moisture and vegetation cover will differ.

Development of a mobile powered hole digger for orchard tree cultivation using a slider-crank feed mechanism gives another example of the design of an auger design, which doesn't straight up work for this case but gives some insights and can be used in this design (Wangyuan, et al.,2016)[20]. The article gives a great example of calculating important factors of the auger. Noticable is the feed rate, being the distance the auger goes down per full rotation. Later in the article the power consumption from the drill is tested experimentally and linked back to factors like the feed rate. This auger had a unique design with a crank influencing the speed at which the auger moved up and down. This could be simplified in the design of our robot by making the auger move down at a linear pace.

An auger experiences certain loads during drilling. Conventional methods simplify soil cutting to mass points, which is not conform reality. Other parameters such as soild pressure variability over the screw haft and the spiral angle also can affect the force distribution. The model described in Mechanical model of hollow-external-screw drill rod for lunar soil particle vertical conveying gives a good insight in the forces (Wei, Wang and Liu, 2013) [21]. While the model is written for lunar soil changing the parameters to earth soil counterparts will make it true for earthly soil

Contemporary considerations regarding reforestation

Next a branch of the general literature review focused on the problem (reforestation) instead of the possibilities for the potential product (robot). Inquiries were made into the scale for the need for reforestation, the involved methodologies and conditions, and lastly the employment of robotics technology in reforestation practices to assess whether there is potential for improvement of the current technologies or if there even are any current technologies at all. The results of this research can be found below.

Reforestation and Forest Fires

Forest fires can be a pain in the ass for national parks such as the Yellowstone National Park. A wild fire will always cause a change in vegetation between the pre- and post-fire situation, which is expressed as the burn severity. Preferably for the regrowth of a forest in a national park the burn severity should be as low as possible. However, for some species the burn severity is much higher than for others, this is primarily the case for vascular species, as freshly burned areas create a situation ideal for the growth of pine seedlings and can hence create a paradigm shift in the decomposition of the forest after a wild fire occurred. Another important factor is the size of the wildfire, as a larger burned area gives a larger potential for tree seedlings of a certain species for which the conditions are favourable to sprout, which will lower the overall species diversity. Hence not every forest fire will generate the same biotic reaction (Turner, M.G. et al. 1997) [22].

In recent years a lot of deforestation has occurred in Latin America and the Caribbean. Then again, a lot of forest recovery has also sprouted, partly caused by demographic and socio-economic change. This is the main factor influencing changes in wood growth. Woody vegetation change was dominated by deforestation in 2001-2010 with a loss of 542 thousand km2, but 362 thousand km2 was recovered due to reforestation efforts, leaving only a net loss of 180 thousand km2 of forest. Considering the eminent dependence of woody vegetation on the dynamics of deforestation and reforestation, these processes which are each other counterparts need to be regulated more extensively (Aide, T.M. et al. 2013) [23].

It is also possible for invasive species to become the dominant factor in forests after a wildfire, this results in a new kind of forest that has a less healthy ecosystem which might even spread to unaffected areas in its vicinity, thus contaminating the rest of the forest. In general, invasive species have a higher survival rate than the species which consisted the original vegetation in the area. Invasive species reproduce faster and their seeds are carried to areas less affected by wildfires. Since the survival rate is relatively high, it is beneficial to remove the leftover seeds that survived the wildfire as to eliminate any potential hostile takeover of the vegetation population since the goal of reforestation in most National Parks is restoration of the pre-fire situation as best as possible (Zouhar et al., 2008) [24].

Current deforestation and combat methods

Deforestation is clearing Earth’s forests on a massive scale, often resulting in damage to the quality of land. The world’s rain forests could completely vanish in a hundred years at current rate of deforestation (if no reforestation effort are made, but only losses due to deforestation, whether they be of human or natural origin, are taken into account). Consequences of deforestation are the loss of habitat for millions of species and climate changes. The most feasible solution to deforestation is to carefully manage forest resources by eliminating clear-cutting to make sure forest environments remain intact. The cutting that does occur should be balanced by planting young trees to replace older trees felled. The number of new tree plantations is growing each year, but their total still equals a tiny fraction of the Earth’s forested land (National Geographic, 2015) [25].

Rehabilitation of deforestation areas can have different steps. It can include anti-erosion works, projects for slope formation and protection and reforestation. The robot-to-be-designed will be focussed on reforestation, with preferably some potential for modularity to increase the number of fields it could be employed in. The forest service takes into account the type of vegetation that has been burned, the success potential of natural regeneration of trees and the general conditions, and, accordingly, shall proceed, or not, to artificial reforestation of burnt areas using native species. The purpose of reforestation is the creation of new forests, the renewal of mature forests and the recovery of degraded forest ecosystems while ensuring natural regeneration or artificial intervention (seeding or planting) for production purposes and the protection of soils. The cost of reforestation in the last 8 years was enormous due to many manhours (Christopoulou, 2011) [26].

However, there is not one clear-cut way to tackle reforestation, as many methods exists and the choice for which method will be used is largely dependent on the context of the problem (e.g. scale, budget, type of forest, etc.) (David, 2015) [27]. Almost all methods have in common that they are based on man work, people are physically present and are planting the seeds themselves: this is known as direct seeding. One method that is currently used that does not involve a person performing physical labour in the planting process is called aerial seeding. This method plants new seeds using planes, helicopters and more recently even drones (BioCarbon Engineering, 2018) [28] (Droneseed, 2018) [29]. This method is much more efficient than being physically present on the ground but is generally outside the budget of most reforestation projects.

Seeds of different species have different optimal depths for sowing, with some growing best if they are buried a few inches deep in the soil, while others, including many grasses and herbs, need exposure to light to germinate and so need to be on the surface. A rule of thumb when growing vegetables and grains is to sow the seed at a depth of one to two times the width of the seed. If seeds of one species, or a mixture of seeds of different species with different needs are randomly mixed in a larger seed ball, at least some of the seeds should be in the optimal position for germination, which optimises reforestatoin (Goosem & Tucker, 2013)[30].

Reforestation also allows for augmenting the composition of the forest, as species can be either suppressed or promoted in the new area. This can result in a healthier forest and allow for a more beneficial ecosystem for animals, but if done wrong it could also be disastrous by introducing a fast-growing undesirable species, such as weed. Thus, some degree of precision is required when replanting the forest along with a degree of vigilance not to include the wrong seeds in the batch of seeds which will be dispersed, as a new composition might result in a new dominant species and therefore a different forest as before. Hence precision is needed to assure certain plants might (not) dominate the forest in certain areas (Yao et al., 2016) [31].

Current use of Robotics Technology in seeding/reforestation activities

The use of machinery in agriculture, the logging industry and nature upkeep is commonplace, however the application of (autonomous) robotic technology is still rather in its infancy. Some robotics solutions exist in these field, which are primarily categorised in 2 classes: a mobile robotic class and a drone class. Examples in the mobile robotic class include the R-Stepps project to combat desertification (Mohamed, Flavien & Pierre, 2015) [32] and the Agribot to plant seeds on farming land (Pavan et al., 2017) [33]. Examples in the drone class include the Treek'lam (Sinalkar & Phade, 2016) [34] and the quadcopter designed by Fortes (Fortes, 2017) [35].

Overall this leaves us with almost countless possibilities for either designing a new robot or improving the existing version of the mobile robot and/or drone.

General conclusions

From the conducted general literature review some general conclusions can be made. Deforestation is global phenomena with drastic consequences for mankind and nature, however there exist effective methods to combat this such as reforestation. Within reforestation different applications exist to get the job done, however these application differentiate in costs, effectiveness and time consumption. Robotics technology is hardly used in this area and is still in its infancy, leaving much room for innovation and improvement. Within robotics technology a promising solution for this problem would be a mobile seeding robot, which equipped with sensors can determine the fertility of the ground and plant the seeds most optimally for regrowth. Preferably this robot would be able to operate autonomously so that it can cover an area (which will be predetermined due to the fire) without the need of supervision of forest rangers which will be busy in the periods after a wildfire. Additionally a preference would be for this robot to be modular so that there is no need to develop a completely new robot for a different task, as it is most likely National Parks will have the need for a robot which can carry out other tasks for the reforestation process such as the removal of seeds of surviving species which might become dominant after a forest fire. However, room for debate still exists upon which mechanism would be best to plant seeds and whether a robotics technology is indeed a desired solution when compared to the contemporary alternatives.


  1. 1.0 1.1 Farritor, S. & Dubowsky, S.. Autonomous Robots (2001) Volume 10, pp57-65. “On Modular Design of Field Robotic Systems”.
  2. Bawden, O., Ball, D., Kulk, J., Perez, T., & Russell, R.. Australian Conference on Robotics and Automation (2014). “A lightweight, modular robotic vehicle for the sustainable intensification of agriculture.” Retrieved from: Accessed at 02-05-2018.
  3. Bak, T., & Jakobsen, H.. Biosystems Engineering (2004), Volume 87, pp 125-136. "Agricultural robotic platform with four wheel steering for weed detection.".
  4. Li, G., Urbina, R., Zhang, H., & Gomez, J. G.. International Conference on Advanced Mechatronic Systems (ICAMechS) (2017), pp 145-150. “Concept design and simulation of a water proofing modular robot for amphibious locomotion.”. IEEE.
  5. Cordes, F., Bindel, D., Lange, C., & Kirchner, F.. Proceedings of the 10th International Symposium on Artificial Intelligence, Robotics and Automation in Space (iSAIRAS’10) (2010), pp. 38-45. “Towards a modular reconfigurable heterogenous multi-robot exploration system.”
  6. Unsal, C., Kiliccote, H., & Khosla, P. K. (1999, August). “I (CES)-cubes: a modular self-reconfigurable bipartite robotic system.”. In Sensor Fusion and Decentralized Control in Robotic Systems II (Vol. 3839, pp. 258-270). International Society for Optics and Photonics.
  7. Murata, S., Yoshida, E., Kamimura, A., Kurokawa, H., Tomita, K., & Kokaji, S. (2002). “M-TRAN: Self-reconfigurable modular robotic system.” IEEE/ASME transactions on mechatronics, Volume 7, pp431-441.
  8. Jorgensen, M. W., Ostergaard, E. H., & Lund, H. H. (2004, September). “Modular ATRON: Modules for a self-reconfigurable robot.”. Intelligent Robots and Systems, 2004.(IROS 2004). Proceedings. 2004 IEEE/RSJ International Conference on (Vol. 2, pp. 2068-2073). IEEE.
  9. Baca, J., Hossain, S. G. M., Dasgupta, P., Nelson, C. A., & Dutta, A. (2014). “Modred: Hardware design and reconfiguration planning for a high dexterity modular self-reconfigurable robot for extra-terrestrial exploration.” Robotics and Autonomous Systems, Volume 62, pp 1002-1015.
  10. Yim, M., Roufas, K., Duff, D., Zhang, Y., Eldershaw, C., & Homans, S. (2003). “Modular reconfigurable robots in space applications.”. Autonomous Robots, Volume 14, pp 225-237.
  11. Hashimoto, K., Matsuo, E., Kawagoe, H., & Iwaasa, H. (1996). U.S. Patent No. 5,554,980. Washington, DC: U.S. Patent and Trademark Office. Retrieved from: Accessed at:06-05-2018.
  12. Kessler, A. M. (2015). "Elon musk says self-driving tesla cars will be in the us by summer.". The New York Times, B1. Retrieved from: Accessed at 06-05-2018.
  13. Matsushiro, Y. (1984). U.S. Patent No. 4,457,101. Washington, DC: U.S. Patent and Trademark Office. Retrieved from: Accessed at 06-05-2018.
  14. Bagloee, S. A., Tavana, M., Asadi, M., & Oliver, T. (2016). Autonomous vehicles: challenges, opportunities, and future implications for transportation policies. Journal of Modern Transportation, 24(4), 284-303.
  15. Barnes, E. M., Sudduth, K. A., Hummel, J. W., Lesch, S. M., Corwin, D. L., Yang, C., ... & Bausch, W. C. (2003). Remote-and ground-based sensor techniques to map soil properties. Photogrammetric Engineering & Remote Sensing, 69(6), 619-630.
  16. Kuang, B. Y. (2012). On-line measurement of some selected soil properties for controlled input crop management systems. Retrieved from: Accesed at 05-05-2018.
  17. Correia, S. F., Antunes, P., Pecoraro, E., Lima, P. P., Varum, H., Carlos, L. D., ... & André, P. S. (2012). Optical fiber relative humidity sensor based on a FBG with a di-ureasil coating. Sensors, 12(7), 8847-8860. (2012),
  18. Martínez Zabala, L. M., Celis García, R., & Jordán López, A. (2014). How wildfires affect soil propierties. A brief review. Cuadernos de Investigación Geográfica, 40 (2), 311-331.
  19. Atondo-Bueno, E. J., López-Barrera, F., Bonilla-Moheno, M., Williams-Linera, G., & Ramírez-Marcial, N. (2016). Direct seeding of oreomunnea mexicana, a threatened tree species from southeastern mexico. New Forests, 47(6), 845-860.
  20. Wangyuan, Z., Jingliang, W., Xiaomao, H., Dong, Y., Yingbiao, Z., & Graham, S. (2016). Development of a mobile powered hole digger for orchard tree cultivation using a slider-crank feed mechanism. International Journal of Agricultural and Biological Engineering, 9(3), 48-56.
  21. Wei, C., Wang, H., & Liu, T. (2013, June). Mechanical model of hollow-external-screw drill rod for lunar soil particle vertical conveying. In Control and Automation (ICCA), 2013 10th IEEE International Conference on (pp. 1240-1245). IEEE.
  22. Turner, M. G., Romme, W. H., Gardner, R. H., & Hargrove, W. W. (1997). Effects of fire size and pattern on early succession in Yellowstone National Park. Ecological monographs, 67(4), 411-433.[0411:EOFSAP]2.0.CO;2
  23. Aide, T. M., Clark, M. L., Grau, H. R., López‐Carr, D., Levy, M. A., Redo, D., ... & Muñiz, M. (2013). Deforestation and reforestation of Latin America and the Caribbean (2001–2010). Biotropica, 45(2), 262-271.
  24. Zouhar, K., Smith, J. K., & Sutherland, S. (2008). Effects of fire on nonnative invasive plants and invasibility of wildland ecosystems. In: Zouhar, Kristin; Smith, Jane Kapler; Sutherland, Steve; Brooks, Matthew L. Wildland fire in ecosystems: fire and nonnative invasive plants. Gen. Tech. Rep. RMRS-GTR-42-vol. 6. Ogden, UT: US Department of Agriculture, Forest Service, Rocky Mountain Research Station. p. 7-32, 42. Retrieved from: Accessed at 10-05-2018.
  25. National Geographic. (2015, April). Deforestation. Retrieved from National Geographic: Accessed on 12-05-2018.
  26. Christopoulou, O. G. (2011). Deforestation/Reforestation in Mediterranean Europe: The Case of Greece. In Soil Erosion Studies. InTech.
  27. David. (2015, January ). Reforestation Methods Reforestation Projects. Retrieved from: . Accessed at 12-05-2018.
  28. Retrieved from: Accessed at 14-05-2018.
  29. Retrieved from: Accessed at 14-05-2018.
  30. Goosem, S., & Tucker, N. I. J. (2013). Repairing the rainforest. Wet Tropics Management Authority and Biotropica Pty. Ltd. Cairns, Australia. ISBN 978-1-921591-66-2
  31. Yao, J., He, X., He, H., Chen, W., Dai, L., Lewis, B. J., & Yu, L. (2016). The long-term effects of planting and harvesting on secondary forest dynamics under climate change in northeastern China. Scientific reports, 6, 18490.
  32. Mohamed, Z., Flavien, V., & Pierre, B. (2015, October). Mobile robotics for restoring degraded ecosystems. In Global Humanitarian Technology Conference (GHTC), 2015 IEEE (pp. 273-278). IEEE.
  33. Pavan, T. V., Suresh, R., Prakash, K. R., & Mallikarjuna, C. (2017). Design and Development of Agribot for Seeding. Retrieved from: Accessed at 05-05-2018.
  34. Sinalkar, S., & Phade, G. (2016, December). Treek'lam. In Global Trends in Signal Processing, Information Computing and Communication (ICGTSPICC), 2016 International Conference on (pp. 611-614). IEEE.
  35. Fortes, E. P. (2017). Seed Plant Drone for Reforestation. The Graduate Review, 2(1), 13-26. Retrieved from: Accessed at 08-05-2018.