====== Fausto Vega REU 2019 Journal ====== **Author:** [[unlv_vega|Fausto Vega]], Email: \\ **Date Last Modified:** 07/20/2019 \\ ===== Week 7 ===== ==== Photo of the Week ==== {{ :vega:img_0919.jpg?upright }} This is a picture of another robot that they have in the lab. It is the Nao robot which is similar to the Darwin OP that we have in the lab. In my opinion, the Darwin is better because of the MX-28 dynamixel actuators. Well I haven't seen any demo's with the Nao because none of them are currently working. They usually just turn it on during demo's and it has a default program that makes it keep eye contact with anyone in range. Something that I have noticed is that during the summer, the Socially Assistive Robotics Group Lab (SARG) has tours every week. Usually the tours consist of orientation students or summer camp students. Yet, they do not have any demos prepared for the students which makes half of them uninterested during the tour. I believe that DASL has a good set of demos that can impress students that want to get into robotics. ==== Weekly Progress/ Next Week Plans ==== This week I have created a program that creates a marker on RVIZ on the distance of picture frames. The YOLO object detection model was trained by my Phd mentor to track people and picture frames. This will let the robot know the distance of the frames in an art gallery scenario. My mentor also wanted to do a context classifier on the robot as he trained a model to decipher between hallways and an art gallery. Yet, I was having problems installing Keras (the machine learning program they use over here) on my computer. There was something wrong with the pip install and I did not want to update it because I heard it messes up existing programs. I also looked into the navigation stack of the robot and how to set it up. I watched several tutorial videos to get introduced to all the nodes amcl, move base, map server, ect as well as all the maps associated with the navigation. I configured the launch files and got a default map and urdf model of the robot to show up on rviz. Now the step that is left is to write the transforms between the robot, laser, and the map. Another goal is to create a new map using the gmapping package. This Monday I met with Dr. Harris' graduate student Nick. He showed me the cool project that they are working on which is the Unity interface for bridge inspection. He said that the interface is connected to ROS, therefore it can send command velocities to the robot. They are currently testing it with a Roomba robot. He also talked about his work with the project. He is focused on using genetic algorithms in order to optimize the coverage of the robots on the bridge. This genetic algorithm creates random paths and then selects the best ones and then repeats the process until an optimal solution is found. They currently have the simulator with Dr. La's bridge robot and a drone. Another idea that he shared with me is that he heard someone else in the Inspire project is working on a drone to pick up the wheeled robots which will overall improve the optimization as the robots will not have to cover certain paths to get to their starting points.He also shared with me a couple papers that describe the project as well as the github repository. This interface also works with VR. Linked below is the github repository. github: https://github.com/simingl/bridgeinspection I also learned that github is a huge resource for code documentation and we should start implementing it in the lab. It will reduce the problem of losing code if a robot ends up failing. It is also a good source of documentation for anyone that wants to develop their own project off of existing code. It is also useful for other people to help debug and collaborate on a project. {{ :vega:img_0920.jpg?upright }} ==== What I Learned about Myself ==== I learned that I get stressed with computer science. I just don't know why certain things fail at certain times. For example, we are using ssh to control the raspberry Pi on the pioneer, and it was working perfectly fine before lunch. Then after I come back from lunch, it was not able to connect to the host or even ping the ip address of the raspberry pi from my laptop. I was bringing back memories from the Las Vegas Geekpwn competition haha. I shared this experience with my graduate mentor and he said that's how robotics is. A huge part of improvement is how robust a system really is. Overall I spend the work day on the computer debugging code which is something that takes patience (and I am learning that along the way). Another thing I learned about myself is that I prefer having certain deadlines such as a WBS. I believe it is more efficient for getting tasks done. The organization of a project is key, and if it is not there, time is wasted with coming up with the next step. Working along a mentor is also efficient as they are there with any questions regarding the project. ==== New Person That I Met ==== This week I met Will, another REU student. He is studying mechatronics at the University of North Carolina Asheville. He told me about his work with a robotic arm to play jenga that they are working on. It sounded like an interesting project that incorporated ROS, machine learning, and computer vision in one.