Autonomous Robotic Defense System

Throughout the history of engineering, the improvement of defense systems of the most diverse nations has been one of the greatest motivations for technological innovations and for the applications of technical knowledge. Whether in rudimentary or sophisticated tools, the development of these technologies is of profound importance for the future of a given group.

With this in mind, the Autonomous Robotic Defense System project aims at the development of an effective defense system, automated and on a smaller scale.

Being in its essence a prototyping project, the knowledge required for its execution varies from mechanical design concepts, material specifications, components determination and physical integration of the solution, to electronics, the development of circuits and embedded systems, and programming, which fits as one of the skills that most differentiate professionals lately, and will be responsible for the system control algorithm. This project offers a great opportunity for training in the various areas of engineering, because problems of reasonable complexity will be solved using simulation, modeling and programming tools.

The project aims to build an apparatus capable of firing a projectile within a radius of 5 meters, being able to hit any point within the range. The figure above represents the problem: The machine that will be developed, in red, must be able to hit any point of the blue circle, for example, the target in yellow. The requirements that the machine must meet are:

- There is no restriction about the firing method, it can be either like a cannon or a catapult;

- The target selection can be manual, by the user, or autonomous, through sensing and communication;

- Use a microcontroller, like Arduino;

- As an extra item of the project, the system can be transformed into a mobile robot, being able to approach the target and make the shot even if it is initially out of range;

This will require several steps of development involving some initial studies, the target detection, the shooting mechanism, the locomotion system and final prototyping.

The project can be divided into three pillars, detection, shooting and movement, the first step being responsible for the initial theoretical foundation of the different possible approaches and which will be adopted for the final solution. The starting point was the choice of a microcontroller and the interface with which the user will communicate in order to control the system, for this were chosen:

• Arduino Mega 2560, for the higher processing power and pin availability;
• Android App, which using bluetooth, through the HC-05 module, will enable communication between the user and the microcontroller;

For the detection was needed a method tolerant to the adverse conditions of the environment and that could distinguish the target from normal objects, being able to detect it with precision in 5 meters. The studies involved:

• Ultrasonic sensors, unable to differentiate the target from normal objects and with a maximum range of 4 meters, with precision losses;
• Infrared sensors, sensitive to electromagnetic interference sources such as the sun and fluorescent lights, but with a range of up to 35 meters when using high power LED;
• LIDAR sensors, high dimensional precision and a range of more than 10 meters, but with a cost 40 times higher than the previous alternatives;
• Servo motors and stepper motors, to enable the movement of the sensor, covering the entire perimeter of the defense system;

The chosen method for detecting targets consists of transforming them into IR Beacons, with LEDs emitting a modulated source of light in 38kHz. The modulation differs the target from ambient radiation and perturbations from other objects, being easily captured by an IR receiver working on the same frequency. To encompass the whole field of view will be used a Servo motor for the vertical movements and a Stepper Motor for the horizontal ones.

Considering that the sensor data is binary, informing only if the target is or isn’t in a certain direction, it is not possible to make angle compensations depending on the distance it is at, for this the projectile must have enough energy to make a trajectory with a minimum drop of altitude. The viable solution found for the problem is the implementation of the concepts from tennis ball launchers, primarily made by:

• High rotation pulleys, where two rotating pulleys convert their moment of inertia to the projectile, in the form of linear velocity, being a mechanically simpler approach when compared with pneumatic systems or airsoft shooters;

To finalize the initial studies for the project the adopted locomotion system was defined:

• Omnidirectional movement, this method allows movements in any desired direction without the need to turn the wheels, even rotations on the own axis;

Although it is an additional requirement, the capacity of locomotion will add in the final result making the system more practical and with more similarity to real defense systems, besides allowing the horizontal movement detection to be implemented by the project's own wheels, eliminating the need an additional motor. NEMA 17 Stepper motors with A4988 drivers will be used.

The first step in implementing a target detection with the infrared receiver was to understand how the infrared communication protocol works, what the modulation steps are and the signal pattern used for sending and receiving data.

This signal modulation consists of an encoder that converts binary data into a series of electrical pulses that change state, between LOW and HIGH, according to a guide frequency. Being the 38kHz frequency the most used due to the rarity of natural sources, making it easily distinguishable from background noise. These pulses are then sent to a LED transmitter which converts them into light pulses, the receiver then converts them back to a binary signal, which will then be sent to the microcontroller. It is worth noting that the photodiode present in the receiver detects all the frequencies of the infrared spectrum, but a filter passes only the 38kHz range, which is then amplified to the supply voltage.

The standard with which the receiver interprets the infrared signal for binary data is defined by means of a transmission protocol, the NEC protocol being one of the most used in projects with the Arduino microcontroller. An example of how this occurs would be the conversion of this signal into a binary "1".

The digit "1" starts with a 38kHz HIGH pulse with 562.5.5µs duration followed by a LOW pulse with 1687.5µs. The digit "0" is transmitted by a 38kHz HIGH pulse with 562.5.5µs of duration followed by another LOW pulse with the same duration. This way it is possible to encode and decode any binary data in the NEC protocol, differing from the other protocols only by the individual duration of the HIGH and LOW pulses.

Therefore, every time a device sends information via infrared, such as pressing keys on a remote control, unique hex codes are generated based on the sequence of bits received, and the microcontroller has the responsibility to process and interpret these codes.

The standard with which the receiver interprets the infrared signal for binary data is defined by means of a transmission protocol, the NEC protocol being one of the most used in projects with the Arduino microcontroller. An example of how this occurs would be the conversion of this signal into a binary "1".

The digit "1" starts with a 38kHz HIGH pulse with 562.5.5µs duration followed by a LOW pulse with 1687.5µs. The digit "0" is transmitted by a 38kHz HIGH pulse with 562.5.5µs of duration followed by another LOW pulse with the same duration. This way it is possible to encode and decode any binary data in the NEC protocol, differing from the other protocols only by the individual duration of the HIGH and LOW pulses. Therefore, every time a device sends information via infrared, such as pressing keys on a remote control, unique hex codes are generated based on the sequence of bits received, and the microcontroller has the responsibility to process and interpret these codes.

The receiver used was the TSOP4838 and it was wired according to its datasheet, and for protecting the signal from power fluctuations a capacitor and resistors were used in series and parallel. The signal reception and emission codes were based on the IRremote library, the most documented Arduino library in this area.

The first idea for an emitter circuit was based on a 555-timer fine-tuned to 38kHz driving some TSAL6200 LEDs, but when tested the IR receiver couldn’t recognize the signal because it was a continuous source of modulated light and not an intermittent one, such as remote controls. With that in mind the final approach used an Arduino Nano, for its small size and ease of use, running the code IRsendDemo from the same library. The output signal of the microcontroller was connected to a BC547 transistor that then drives the LED with the appropriate current. The tests resulted in a detection range of more than 5 meters with a precision of around 2 degrees.

The servo motor for the Tilt movement was the MG996r because of its metal gears and budget price. As it consumes around of 2,5A in 5V, its power will come from an external supply to avoid overloading the voltage regulator in the Arduino.

A NEMA 17 Stepper motor was designated for panning the receiver, considering that the same motors will be used later for locomotion. A988 drivers will be used to control them, where each pulse sent to the STEP pin of the drivers result in a 1.8-degree rotation.

All the circuits of the project were designed using the online tool EasyEDA.

The structure of the code was built modularly making it easily modifiable as new features are added to the project. It’s based on functions that are executed depending on the mode that the user chooses, between manual and autonomous. In the manual routine the user enters the values for the angle and inclination of the target and after the motors reach the desired position the shooting protocol begins. For the autonomous section the movements of Pan & tilt are synchronized with the reception of signals, when the target is found the current position is saved, triggering the shooting protocol too.

To validate the effectiveness of the target detection a device was modelled that enables the Pan & Tilt movement and accommodates the IR sensor. The 3D modelling was made with SolidWorks, based on reference geometries from GrabCAD, and 3D Printed in PLA for 5 hours.

The initial tests of this prototype were performed with different power supplies among the components, generating a lot of electrical interference and making it almost impossible to use the circuit. Later a 12V battery was adopted as a single power supply and with common GND, solving the problem. The servo motor cables were also wrapped in an iron core to reduce electrical interference by capacitive coupling, that could affect the infrared detector, already externally isolated by a layer of insulating tape and aluminum foil.

After these changes the prototype was able to accurately detect targets at distances of up to 10 meters, validating the detection concepts.

The method chosen for the firing mechanism involves two pulleys in high rotation, however for its development it was necessary to specify, with assertiveness, the motors, the projectiles and the pulleys used. In order to substantiate this specification, the launch path was modelled mathematically, considering the effect of the air resistance on the projectile in a parabolic path. To model the drag force the following equation was used:

Regarding the initial parameters, we used the standard value of the air density, according to IUPAC, of 1,225 kg/m3 and the drag coefficient of a sphere of 0.47. For the cross-sectional area, it is necessary to know the dimensions of the projectile, so several candidates were considered, such as bearing balls, lead balls and airsoft balls. The projectile chosen were airsoft balls, also known as BBs, due to their low cost, high dimensional accuracy and low probability of damaging the infrared target. Considering that their radii are 3mm, this results in a frontal area of approximately 0.00002827 m2.

With that in mind, the first step was to mathematically model the trajectory as a parabolic path considering the drag from the air. It was used an Excel table created by the Real-World Physics Problems site that calculates the projectile movement in respect to some initial parameters, like the air density, drag coefficient and cross-sectional area of the projectile, and output a graph. After some considerations it was concluded that airsoft BBs were the most suited projectiles for this application due to their low cost, high dimensional accuracy and low weight, and so the appropriate parameters were included on the formula.

By comparing the graphs, it is possible to conclude that if the shooter accelerates the projectiles to speeds between 50m/s to 60m/s they will suffer a drop of less than 5cm in 5 meters, being enough energy to disregard altitude compensations during the shots.

Knowing that each pulley will contribute with half of the linear velocity and considering an initial radius of 40mm, subject to changes, the tangential velocity formula of a disk returns an angular velocity, by pulley, of at least 750 rad/s, or 7161rpm.

Considering some factors like the low power consumption, compact dimensions and an operational voltage below 12V, two motors were chosen to make some tests: A DVD motor, DC RF-300EA-EH and a 130-Size Motor, very common in child toys.

As in their nominal voltages both did not reach the desired rotation for the pulleys, the formulas of the constants of an electric motor and the relation between torque and speed were used, together with the data of their datasheets, to calculate what is the necessary voltage and developed torque in order to reach rotations above 7000rpm.

Graphs were generated in the online tool Desmos that show that fed with 8V, the RF-300 motor can dissipate up to 22% of its torque in friction and still maintain a speed of 7000rpm, and fed with 9V, the 130 motor can dissipate up to 48% of its torque still reaching the desired speed. The choice between the two motors was made during the physical prototyping of the shooter, where the precision and range results were the determinant. For the tests, the power used was supplied by a buck converter with variable voltage and controlled by an Arduino by the activation of a relay module.

The prototypes were totally 3D printed, and had the objective of testing the shooter concept, as well as the coupling tolerances of the motors and the drag of the projectile. The starting point was the first version of the pulley, with 40mm radius, higher mass concentration at the edges and a central hole for motor coupling.

With this in mind the first shooter was modelled and tested.

The tests showed that it did not guarantee a repeatability of the shots, either by the lack of contact of the pulleys with the projectile, or in the variations introduced when manually feeding the BBs, besides that there is not a closed barrel to guide the trajectory and increase the precision.

The next prototypes aimed to solve the problems encountered in the first version, varying one dimension/parameter at a time and analyzing the results, such as the distance between the pulleys, the length of the barrel, the diameter of the barrel and the height of the feed pipe, which was added later.

After reaching a stable platform for testing, the results of the motors were compared, demonstrating a better performance of the RF-300 ones, due to a considerable speed variation between the two 130 motors.

Therefore, the DVD motors were chosen for the next iterations of the shooter platform.

In total 12 versions of mechanisms and 14 versions of different pulleys were designed and printed, so that the final version could meet the initial requirements of range and accuracy, this version has:

• An inner barrel of 6.3mm in diameter and 90mm in length, printed separately for best dimensional accuracy;
• 39mm radius pulleys with 2mm rubber edges, and a total thickness of 6mm, ensuring a uniform contact area with the projectile;
• A 30mm high feed pipe to standardize the velocity of projectile entry;
• A lower support structure, with the purpose of reducing the device's vibrations, stabilizing the motors and providing a fixation point for the end of the barrel;

Being capable of reaching distances above 10m with ease, and for distances of 5m the average grouping was 4cm around the target, in accordance with the mathematical modeling and sufficient to eliminate the targets detected by the system.

In order to develop an autonomous defense system, a device capable of precisely and autonomously feed the projectiles is needed.

Therefore, to integrate this functionality into the final version of the shooter, a method was developed using a gear where the spacing between the teeth corresponds to the silhouette of the BB. This gear remains enclosed with only a lower opening for the feed pipe and another for an upper magazine. To rotate it was used a 28BYJ-48 stepper motor and un ULN2003 driver. The gear has a capacity for 9 projectiles, so for every 40 degrees rotated a projectile will be carried from the magazine to the feed pipe.

The motor and driver will be powered by 5V and connected to the microcontroller by 4 digital pins. For the control code the library used was the Stepper, where for each execution of the code the motor moves 40 degrees.

The shooting protocol was organized in a function called disparo() so that it can be implemented directly in the final code, such as the detection. The logic consists of enabling the relay module, to turn on the motors, and then rotating the stepper motor in 40 degrees, process that gets repeated each time the code runs.

Practical tests of the coupling between the final prototype of the shooter and the automatic feeder showed that a lot of vibration is introduced into the system by the feeder not being fixed to the body of the shooter at other points. That is, for the final prototype of the defense system a more robust coupling should be developed and the use of insulating materials such as foams or rubbers considered.

In terms of precision, the mechanism's operation does not present obstacles such as clogging or flaws in the transport of the projectiles, so an optimization of the geometries is not necessary.

The desired movement is the result of the combination of the force vectors exerted by each wheel, a fact previously commented during the planning, however to understand what is the explicit relationship between the angular velocities of each wheel and the linear velocity of the system as a whole and what are the variables necessary to control it the mathematical analyses present in the paper Modeling and Kinematics Simulation of a Mecanum Wheel Platform in RecurDyn will be used. This paper aims at the development of a kinematic simulation of a Mecanum Wheel Platform, thus detailing the process of obtaining these variables.

The kinematic model applied in this project is based on a typical 4-wheel longitudinal symmetrical structure, as shown in the diagram above. The coordinate system is aligned with the center of the platform, the angle of the bearings is 𝛼 and assuming that there will be no loss of grip with the floor, one can establish the kinematic equation as showed above. And considering alfa as 45 degrees, the system with those 4 equations is obtained.

Therefore, having the knowledge of the velocity vector components and the angular speed makes it easy to convert it to the individual engine rotations, thus enabling the implementation of a control algorithm based on the data that the user sends. With this in mind the development of a communication interface between the user and the microcontroller is the next step for the effective movement of the system.

The Android appl that composes the base of the system communication was developed by the software MIT App Inventor, which will send data by bluetooth to a HC-05 module, connected to the microcontroller. The module's power supply is made by Arduino's 5V, and the communication by RX and TX pins. As the module only accepts 3.3V signals, a voltage divider was used to reduce the 5V logic level of the microcontroller to the acceptable level.

The algorithm used for the application consists of sending a sequence of bytes according to a pre-established timer, where the first one indicates the beginning of the transmission and the following ones represent the values of the system’s variables. The microcontroller synchronizes the reception with the initial byte, storing the received values in a list and addressing them to the corresponding variables. An advantage of this method refers to the ability to add as many values as necessary without changing the final structure.

Regarding the app interface, it consists of two different operation modes, manual and automatic. In manual mode, the user can control the defense system through joysticks, sliders and buttons, allowing a greater interactivity and encompassing all the features of the prototype. For the automatic, the approach chosen informs the current position of the defense system, sent by the microcontroller, as it performs the detection routine. When the target is found the status of the app is changed from "Scanning", to "Target found".

To finish the communication process, the data reception was encoded in the microcontroller, using the logic previously described.

The prototype designed to test the control algorithms used a model of Mecanum Wheels already optimized for 3D printing, which was part of a set of tutorials from How To Mechatronics site. Brackets for the stepper motors were also modelled and printed, wired according to the detection circuit, and later fixed on an aluminum base with the wheels attached. The rollers and fixing brackets were printed using PETG filament, ensuring greater mechanical resistance, and for the external plates of the wheels PLA filament.

In total was 8h and 320g of printing in PLA and 8h and 140g in PETG with a total cost of U$17.48, cost around 2 to 3 times cheaper than the solutions found in the market today.

The code to validate the omnidirectional movement consists of acquiring data from the Bluetooth in the begging of each execution and then updating de velocities according to the Kinematic equations. For the communication was used the SoftwareSerial library, that enables the creation of new Serial Interfaces beyond the common RX and TX pins, that are busy during the upload of new codes. For controlling the velocities, the AccelStepper library was chosen considering the ability to easily change the parameters of each motor and to run more than one at the same time.

The static tests proved the effectiveness of both the app and the algorithm of the microcontroller, by printing the values of the speeds in the serial monitor.

Due to the low current running through the motor coils during the dynamic tests, the system encountered difficulties in performing movements that needed more torque, such as the lateral movement. The solution found was to perform the current adjustment by rotating the potentiometer present in the driver, reaching values of 1.5A after the adjustment, in contrast to 0.6A during the initial analysis. Another modification was the addition of heat-shrinkable tubes around each roller to increase adhesion and improve contact with the ground.

The changes made it possible to perform the most diverse movements without difficulty, but the increase in current in each driver caused an overheating both in the integrated circuits and in the protoboard tracks, causing bad contact and damaging the drivers. Considering these factors was determined the need for an electronic circuit that mitigate contact problems, capable of a better current conduction, active cooling and reliability in its use, therefore the development of a printed circuit board was the solution adopted in the final prototyping stage.

The final circuit was obtained by joining the individual circuits from each section, the objective here is to fabricate a robust PCB (Printed Circuit Board) that works like an Arduino shield with standard connections and support for adding future devices to the project. The power comes from a 5Ah 12V battery and was divided to a 12V line for the Arduino and stepper motors, one 5V line for the drivers and servo motor and one 5V line from the Arduino to power the sensitive components like the IR receiver. Two LM7805 were used to cope with the current requirements. Extra terminals were added to power future components, as well as an ON/OFF switch and jumpers for the driver’s MS1, MS2 and MS3 pins, to change the micro stepping modes.

The method chosen for manufacturing the printed circuit board was by selectively etching the copper layer of a virgin phenolite board using ferric chloride. For this purpose a circuit layout was produced from the final diagram, both in the online software EasyEDA, and then printed on magazine paper using toner.

With the printed layout, it was transferred to the copper board using temperature and pressure of a clothes iron, just enough to soften the toner and fix it on the copper. With the complete transfer, the plate was dipped in the etching solution to start the corrosion.
In total 4 boards were etched and perforated for the final circuit and 3 of these were soldered and tested, until reaching as a result a reliable version that presented no problems regarding the perforation's quality, bad contacts and short circuits between the traces.

Thus the final version of the boad has a heat sink for the voltage regulators, a printed base in PETG coupled to the Arduino to facilitate handling and a layer of super glue on the traces to ensure electrical insulation and mechanical resistance. The board has a peak consumption of around 15A, considering the maximum consumption of each motor, being cooled by a 12V cooler powered by the extra power sockets.

The approach adopted for the final prototype was inspired by the designs of projects such as DARPA's GXV-T, the PL-01 tank and the Robomaster S1 drone from DJI, where through the development of a vehicle platform it is possible to implement the omnidirectional movement, pulley shooter and infrared detection, consolidating it as a practical defense system inspired by already validated industrial applications.

Thus, the modelling of the prototype can be divided into three parts: the shooter, the chassis and the target. One of the objectives for the design is to reuse the maximum number of components already printed during the past steps, besides producing the parts using 3D printing and the advantages, in terms of geometry, that this manufacturing method allows.

The shooter comprises the shooter mechanism, with its feeder, the infrared detector and the fixation to the base, which should allow the tilt movement of the structure as a whole.

The tilt movement was made possible by a gear coupling between the servo motor shaft and the upper structure.

After fixing this coupling on the structure, the shafts fit into two sockets attached to the base of the system together with the servo motor bracket and an external reinforcement for greater mechanical stiffness.

The design for the chassis met certain requirements regarding the mechanical integration of the systems, such as the space for fixing the electronic components, a cooler support and openings allowing the exchange of air with the environment. It was also necessary to accommodate the battery and have a cover that can be removed, giving access to the interior of the system.

After modelling the central part, which was divided into 3 parts to be printed separately, lower skirts were added, fenders for the wheels and the cover with a cooler support.

The aluminum base used in the locomotion section was fixed to the chassis structure, as well as the final details of the cooler and the rear grid.

By joining the shooter to the chassis, the final modelling of the defense system is obtained.

A cover for the emitter circuit was modelled with an extension for the TSAL6200 LED, positioned outside, turning it into a practical and interactive target, based on a disc that when hit is knocked down.

The final integration between the individual codes seemed like a logical evolution, taking account the fact that the previous structures were based on modular functions. One of the characteristics of this new code is the preference on choosing functions that don’t slow or stop the execution of the algorithm, avoiding desynchronizations when receiving data and enabling a smoother control of the system.

The body of the code consists of two main modes, as presented on the Detection:

Inside the manual routine are executed the functions referring to the actuators of the system, like the wheels, the inclination of the mechanism, the gun and the other relay pins. Being constantly updated to encompass all the variations of the variables.

On the automatic mode a detection protocol is used, where the system executes a small horizontal move followed by a vertical sweep, always checking the IR receiver for the target. After the detection, the shooting protocol begins and after it finishes the detection protocol restarts in search of new targets.

The app Interface makes it very easy and practical to control all the options and the functionalities.

After the microcontroller algorithm and the application were finished, and with the circuit board ready, the different parts of the final prototype were printed. As could be observed during the development of the final modelling, 4 main colors were used for the prototype, chosen according to the filaments available during printing. For the silver shades the Silver PLA filament of Good 3D Printing and the Silver PLA of 3DFila, more bluish, were used. The red color represents the PETG Red filament and the white PETG White, both of Good 3D Printing.

A total of 90 hours and 21 minutes of printing were necessary for finishing the 83 pieces that compose the physical prototype, considering also the components printed during the past steps, with a final filament consumption of 1548g, being 809g of PLA, 583g of Red PETG and 156g of White PETG. The final cost of the printed prototype was around U$102.91, considering the material used, the energy consumption, the time and the depreciation of the machine.

In relation to the components needed to finalize the defense system, already accounting for the materials used in the previous steps, the cost was U$71.45, according to the table above.

*and all the components accounted with a final price of zero did not have to be purchased.

After the assembly of the printed parts, the electronic circuit was implemented, a lighting system added and some holes were made to guide the cables of the electronic components and enable the addition of a USB cable to externally power and upload codes to the microcontroller. These cables were organized with plastic spirals, standardizing the external connections of the prototype.

During this step the infrared detector connections were guided separately to avoid interference from the pulley motors, as well as a layer of foil was added to absorb undue radiation.

Another modification to mitigate electrical interference effects was to spiral the servo motor power cables, improving the magnetic compatibility between them and reducing the effects of inductive coupling in the electronic circuit.

With these modifications the integration of the prototype was finalized and the practical validation could be started.

The defense system tests were successful, fulfilling the initial requirements and validating the proposals defined in the study phase. The system offers an almost instantaneous control of the actuators when in manual mode, with movements with great speed and precision, as well as an effective range of 10m with the shooter mechanism and capacity for 19 projectiles. And in automatic operation mode it offers detection accuracy, performing the scanning routines without interference and finding the target with an average accuracy of 2 degrees in 5 meters, managing to knock down the beacon’s disk at any point of the initial radius, even in different heights.

The project of the autonomous robotic defense system was successfully completed and the result obtained in the end was an app-controlled robot capable of carrying out accurate shots, both manually and automatically, on targets arranged in a radius of more than 5 meters, counting on omnidirectional movement to move around the environment and approach the targets effectively. The development of this project allowed a unique training experience in several areas of engineering, such as:

- CAD modelling, using SolidWorks;

- Development and simulation of electronic circuits, using EasyEDA;

- Prototyping of printed circuit boards;

- Android application programming, using MIT App Inventor;

- Programming in C++, using Arduino IDE;

Besides the practical application of concepts in:

- Mechanical projects;

- Projectile firing;

- Communication protocols;

- Embedded electronics;

- Robotics;

With emphasis on the extensive use of 3D Printing in the prototype manufacturing, being essential to enable the production of complex geometries and enable rapid prototyping of various models, resulting in the beautiful final design of the defense system.

This experience has been configured not only as the true extension of STEM studies but as the consolidation of an engineer's and maker’s purpose, using the knowledge to find practical and creative solution for the most diverse problems.

Here is a PDF containing all the reference material and useful links used for the project :)