A HOME MADE ROBOTIC PLATFORM BASED ON THEO JANSEN MECHANISM FOR TEACHING ROBOTICS

Transcription

1 A HOME MADE ROBOTIC PLATFORM BASED ON THEO JANSEN MECHANISM FOR TEACHING ROBOTICS Alejandra C. Hernández, Clara Gómez, Jonathan Crespo, Ramón Barber Robotics Lab. Departamento de Ingeniería de Sistemas. Universidad Carlos III de Madrid (SPAIN) Abstract Nowadays, robotics is becoming increasingly popular as an educational platform. Working on robots provides a unique learning experience. The students receive strong and meaningful feedback for physically experience their work, allowing them to acquire important skills that will help them during their school years and beyond (scientific approach, problem solving skills, creativity). This paper proposes a robotic platform based on Theo Jansen's natural gearing mechanism. This work is focused in the design and development of the low cost robotic platform. The components used are readily available and inexpensive which makes it easy to build by students. The same principle was applied to software, with the focus on open source operating systems, libraries and development tools. With this platform some aspects related to mechanics, electronics, computation, mobile locomotion and surface adaptation can be studied. The design of the robot has been completed with sensors, actuators and a microcontroller that manages all the information. Some modifications in the original mechanism have been needed to get the final prototype. To build the prototype, 3D printing to create the parts of robot has been used, making easier the construction by students. The platform includes a webcam as a sensor in order to allow robotic experiments related to environment modelling and navigation tasks. Finally, to test the robotic platform developed, several experiments have been done. On the one hand, experiments focused on hardware stability and surface adaptation, in which students can learn about robot balance problems. On the other hand, experiments related to environment modelling and navigation that allow the students to work with robot sensors, environment representation concepts and low-level navigation problems. Keywords: Innovation, technology, research projects, robotics, mobile robots. 1 INTRODUCTION Popular interest in robotics has increased at an astonishing rate in the last several years [1]. Robotics technology has been implemented in a variety of fields including medicine, elderly care, rehabilitation, education, home appliances, search and rescue, car industry and more [2]. Mobile robotics is increasingly entering the curricula of many technical studies. Robotics is gaining terrain in industry and consequently more firms are recruiting candidates with experience in robot programming. For this reason, many universities are teaching robotics in their master and degrees programmes [3]. A common approach when teaching robot programming is the use of simulations, in which the user can create different robot configurations. Although simulators are a very useful tool in robotics, allowing to create all necessary robots, evaluate, refine and adjust control programs before taking the real robot, the problem is that things do not work the same way in a real platform. In addition, students do not enjoy the same testing their ideas on a real robot than on a simulated one. Working with real robots opens a new way of teaching robotics, allowing among other things, developing strong knowledge of the fundamentals and experience with real robot systems, give students hands-on experience with real problems, experience in understanding and implementing robotics principles from primary research literature, and give them confidence in their ability and help to develop teamwork skills [4]. Up to now, different types of mobile robots have been used for education [5]. Mainly, these platforms have been wheeled robots [6]. However, legged robots, as the one presented in this work, have also

2 been studied. The education field has drawn attention of biped robots [7] and multi-legged robots ([8], [9] and [10]). This type of robots implies different experiments and practices ranging between stability, interaction, navigation, environment detection, etc. The robot introduced in this paper is a multi-leg robot based on 3D printing technologies. Using this technology increases the scope of this work thanks to its recent implementation in education centres, as seen in other projects [3]. 2 GENERAL DESCRIPTION OF THE PROTOTYPE A multi-leg robotic platform has been developed, which consists of: four legs, two on each side, two servomotors to give movement to the robot, a web cam as sensor for perception and finally an Arduino microcontroller to manage the entire system. Fig. 1 shows the robot built. The prototype is based on the Theo Jansen mechanism, a cinematic artist and sculptor, which build big figures by imitation of animal skeletons that are able to walk using wind force. The structure of the robot has been created using 3D printers. Being a low-cost robotic platform, the components used are readily available and inexpensive which makes it easy to build by students. Fig.1. Developed robot Regarding software, it has been developed an algorithm for topological navigation, performed in the programming language C++ with Open Source Computer Vision (OpenCV) libraries. This legged robot has been designed to walk in different surfaces, taking into account that legs are more effective in an uneven environment than wheels. Likewise, with the developed algorithm based on detection of colours and shapes, the robot is able to manage sensory information in real-time and establish an appropriate navigation. Finally, with this platform some aspects related to mechanics, electronics, computation, mobile locomotion and surface adaptation can be studied. In the following sections, more detailed information about the different aspects related with to the construction and setting up of the robot will be given. 3 HARDWARE In this section all the hardware elements used in the construction of the robot as well as the entire assembly process are detailed. The information is presented into three subsections tackling the structure components including the pieces design, electric and electronic devices used and the platform construction.

3 3.1 Structure components As already mentioned, the design of the legs of the robot presented in this paper is based on Theo Jansen mechanism [11]. This mechanism provides an easy way of simulating the walking of a real leg controlled by a single element which could be a motor or wind. In this case motors are used. The mechanism consists of 7 solids (plus bedplate): 5 solids are binary bars and 2 solids are ternary bars. The bedplate is also a binary bar. In Fig.2 the Theo Jansen mechanism is shown. Solid Nº of joints Joint Bedplate 2 O C 1 2 O A 2 2 A B 3 3 B C D 4 2 D F 5 2 C E 6 2 A E 7 3 E F (- G) Fig.2. Theo Jansen mechanism The shaft to which is attached the motor is at point O. In addition, point C is fixed and is aligned with point O what can be called the chassis or body of the animal or mechanism [12]. Point A is adhered to the circumference or can also go on some region of the circle which generates the limb movement until the lower leg point G describe a movement as shown in the Fig. 3. Fig.3. Trajectory of point G of Theo Jansen mechanism The advantage of this trajectory is that if the ground is uneven, the distance between the mechanism and the ground is short and can be suspended in a better way. This mechanism makes their measures cannot be anything, but must be substantially scaled the original; but it allows small modifications to optimize the trajectory of the point G depending on the steps of the machine.

4 In order to build the robot, the components of the structure have been designed with software for creating solid 3D CAD models called OpenSCAD [13]. Subsequently, 3D printing of the pieces has been made. The printed parts are the legs of the robot, the base for the Arduino and supports for motors. In Fig. 4 and Fig. 5 examples of the design of some elements of the robot are shown. Fig. 4. Design the parts of the robot limb Fig. 5. Foot robot design 3.2 Hardware elements Movement To guarantee the movement of the robot, two servo Futaba S3001 (Fig. 6) have been used. These motors offer good performance for the type of application, in addition to its favorable value. These actuators are used in modelling, so they have the ability to be located at any position within its range of operation, and remain stable at that position. The motor consists of a direct current (DC) motor, a gearbox and a control circuit, and their operation margins are generally less than one complete turn. Fig. 6. Servomotor Futaba s3001

5 3.2.2 Processing Unit The entire process is performed through an Arduino UNO [14], Fig. 7, which is an open-source electronics platform, based on a board with an Atmel AVR microcontroller, with input/output ports and a development environment that implements the programming language Processing/Wiring and boot loader that is executed on the board. It is important to note that to power the Arduino and the servomotors 4 AA batteries connected in series each providing 1.5 Volts are used. Fig. 7. Arduino UNO board Vision Sensor As mentioned, the device used to perceive the environment and receive information is a webcam. Specifically, a Logitech 720p camera (Fig. 8) is used. In Table 1 most important specifications of the camera are presented. Connection type Camera specifications Corded USB USB Type High Speed USB 2.0 Field of View (FOV) 60º Optical Resolution (True) 1280 x MP Video Capture 360p, 480p, 720p, Frame Rate (max) 640x480 Table 1. Features - Logitech 720p The camera (Fig. 8) is positioned in the center of the robot base, and must be connected to the computer via USB to run the navigation algorithm. Fig. 8. Logitec 720p camera

6 3.3 Platform assembly To assembly the robot, the first step is the adjustment of the motors. Selected servos can only be rotated from 0 to 180, so for make continuous 360º turns two actions have been performed: first, disengage the potentiometer which tells the servo the position of the control arm, and then, removal of mechanical stop that prevents the servo rotates over 180 degrees. Subsequently, the assembly of all printed pieces as well as the connection of all electrical and mechanical elements mentioned previously is done. During this phase, several tests in order to adapt the final robot design are performed. First, it started with the idea of printing screws specially designed for the platform; however, due to the very small diameter of the screws it was not possible to achieve the correct printing. So this option was discarded and decided to use metal screws and nuts. Similarly, different tests have been performed with respect to the robot foot and the base where the motors and the board are placed, in order to enhance the stability in the whole structure and improve the grip and the natural foot movement. In Fig. 9 (a) the initial leg assembly is shown, in (b) different foot to the robot are displayed. (a) (b). Fig. 9. Robot assembly: (a) Initial leg assembly, (b)different foot to the robot With the settings made during the assembly stage, the proposed legged robot is obtained, as shown in Fig. 10. (a) Fig. 10. Assembling the robot: (a) with a base, (b) with double base (b)

7 4 SOFTWARE In general, a mobile robot works with a lot of sensory information, which after analysis and postprocessing is used to navigate. This sensory information is subject to the own constraints of the sensors used, including the introduction of small accumulative errors. Therefore, the use of a unique and robust model to these sensory errors for robot navigation is sought. Because of this, using topological information instead of geometrical information associated with the physical position of the elements of the environment has been considered. Through topological maps the characteristics of the environment of the robot and the relationships between them can be represented. Topological maps are composed of nodes and arches. In the case of this project, nodes are formed by the elements of the environment that the robot can find, and arches are the behaviours to be followed by the robot to move through different nodes. In this paper an algorithm based on topological navigation [15] has been designed and implemented. In Fig. 11 the general functioning of the algorithm can be seen. Fig. 11. Flowchart of the general operation of the robot navigation Between the fundamental stages that compose the algorithm is on the one hand, the construction of the navigation map through which the robot has to move. This map consists of the different labels and relationships that the robot can detect both shape and colour in the environment. On the other hand, to carry out the detection of colours and shapes, Microsoft Visual Studio C ++ with the OpenCV libraries [16] have been employed. Through this program in C ++ it seeks to detect the different labels on the stage work to navigate across it. In Fig. 12 the flowchart for detecting colours and shapes is shown. Fig. 12: Flowchart for detecting a specific colour square

8 5 EXPERIMENT DESIGN AND RESULTS The experiments performed offering a friendly environment that could be used by non-experienced people. It is noteworthy that experiments presented below are intended for educational robotics, focusing on teaching the use of new technologies as 3D printers, as well as in learning different aspects such as stability and navigation of legged robots. 5.1 First experiment: different surfaces Three different test scenarios have been designed to study the behaviour of the system on different surfaces. In all of them it has generated a simple map in which each node is located to the left of the previous node. In this way, the robot can find the next node by turning the direction in which it is located. The surfaces used are: cork, wood and granular floor. This type of testing allows the study of the stability problems of the robot, so that the results may determine needed changes in the design of one or more parts of the platform to solve the problems. Through the obtained results, it can be seen that in the wooden floor the robot tends to slip because it is a smooth surface, which does not occur in cork floor, because is softer and rougher than wood. These cause problems of stability that has been improved considerably modifying the feet of the prototype. On the other hand, in a granular surface, that is porous and has level changes, the robot is able to overcome successfully uneven terrain, up and down without much problem than in flat ground. In Fig. 13 the robot during the tests in wooden and cork surface is shown. (a) Fig. 13.The robot during the test in wooden surface (a) and cork surface (b) (b) 5.2 Second experiment: using tags of colours This test consists of detecting the different labels on the environment work to allow navigation of the robot. Each label stores the values of H, S and V fixing the threshold image binarization for proper detection of different colours. As already mentioned, the developed robot navigates based on a topological map in which nodes correspond to different labels of the environment, being the arcs behaviours that the robot must follow to move from label to label. According to this, three different behaviours have been defined (Fig. 14): Behaviour 1: The robot must be directed to a label that is on your right: it has to turn 90º and looking forward corresponding label. Behaviour 2: The robot must be directed to a label that is on your left: it has to turn -90º and looking forward corresponding label.

9 Behaviour 3: The robot has to go to a label that is in front of the current: it has to turn 180º and advance to look for that label. Fig. 14. Example of topological navigation map based on behaviours This type of testing allows become familiar with navigation tasks in robot with legs. In this case, with the use of an algorithm for detection of colours and shapes it is possible to experience and discuss topics related to image processing and object detection. Fig. 15 shows some results of appliying the developed algoritm where the robot detects rectangular colour marks. Fig. 15. Results of the detection algorithm developed 6 CONCLUSIONS In this work, the design and construction of a multi-legged robotic platform as well as, an algorithm for detection of shapes and colours, able to manage sensory information in real time and set the robot navigation properly have been performed successfully. Through the implemented design it has been possible to reduce the complexity of building this type of robots. The algorithms developed, offered as main advantage that has not been necessary to use more sensors that a web camera. This project offers important features for educational purposes. First, it allows students to design new custom pieces simply being printed and tested very quickly. Second, the robot can be thoroughly studied, modified, copied and distributed by anyone. Finally, this project allows entering the world of robotics, and studying the main aspects related to this type of robots, such as stability, navigation, image processing and object detection. As shown in the results, there are many experiments that can be developed using this robot and the user will get a deeper knowledge about the aim of mobile robotics.

10 As future work we plan to continue evolving the robot, making adaptations to the hardware design, incorporating more legs to the robot, as well as developing and improving navigation algorithms and adjustment paths. 7 ACKNOWLEGMENTS The research leading to these results has received funding from the RoboCity2030-III-CM project (Robótica aplicada a la mejora de la calidad de vida de los ciudadanos. fase III; S2013/MIT-2748), funded by Programas de Actividades I+D en la Comunidad de Madrid and cofunded by Structural Funds of the EU and NAVEGASE-AUTOCOGNAV project (DPI C3-3-R), funded by Ministerio de Economía y competitividad of SPAIN. REFERENCES [1] Benitti, F.B.V. (2012). Exploring the educational potential of robotics in schools: A systematic review. Computers & Education. 58: pp [2] Eguchi, A. (2014, July). Robotics as a Learning Tool for Educational Transformation. In 4th International Workshop Teaching Robotics, Teaching Robotics & 5th International Conference Robotics in Education. pp [3] J. Gonzalez-Gomez, A. Valero-Gomez, A. Prieto-Moreno, M. Abderrahim. (2012). A New Open Source 3D-Printable Mobile Robotic Platform for Education. Advances in Autonomous Mini Robots, pp [4] Rawat, K. S., & Massiha, G. H. (2004, April). A hands-on laboratory based approach to undergraduate robotics education. In Robotics and Automation, Proceedings. ICRA' IEEE International Conference on (Vol. 2, pp ). IEEE. [5] Muvin, O., Stevens, C., et al. (2013). A review of the Applicability of Robots in Education. Technology for Education and Learning. [6] C. Gómez, A. C. Hernández, J. Crespo, R. Barber. (2015, November). A ROS-based Middlecost Robotic Platform with High-performance. The 8th annual International Conference of Education, Research and Innovation. Sevilla, Spain. [7] K. Dautenhahna, C. L. Nehaniva, M. L. Waltersa, B. Robinsa, H. Kose-Bagcia, N. Assif Mirzaa, M. Blowa. (2009). KASPAR a minimally expressive humanoid robot for human robot interaction research. Applied Bionics and Biomechanics, vol. 6, pp [8] D. Belter, P. Skrzypczyński, K. Walas, D. Wlodkowic. (2015). Affordable Multi-legged Robots for Research and STEM Education: A Case Study of Design and Technological Aspects. Progress in Automation, Robotics and Measuring Techniques. Vol. 2, pp [9] Metabot: robots in motion. Robotics platform for education, research & fun. (2016). Retrieved from : [10] RobotShop: putting robotics at your service. Hexapod robots. (2016). Retrieved from: [11] Nansai, S., Elara, M. R., & Iwase, M. (2013). Dynamic analysis and modeling of Jansen mechanism. Procedia Engineering, 64, [12] Kim, S. W., & Kim, D. H. (2011). Design of leg length for a legged walking robot based on Theo Jansen using PSO. Journal of Korean Institute of Intelligent Systems, 21(5), [13] Kintel, M., & Wolf, C. (2011). OpenSCAD, The Programmers Solid 3D CAD Modeller. [14] Pomares, J. (2009). Manual de Arduino. Grupo de Innovación Educativa en Automática. Universidad de Alicante. [15] R. I. B. Castaño, Desarrollo de una arquitectura para robots móviles autónomos: aplicación a un sistema de navegación topológica. PhD thesis, Universidad Carlos III de Madrid, [16] Bradski, G., & Kaehler, A. (2008). Learning OpenCV: Computer vision with the OpenCV library. O Reilly Media, Inc.".