Human to Robotic Arm Action Learner
This project is for the pattern recognition lecture which aims to create a deep autoencoder which would map human actions to the UR5 robotic arm. For human skeletal data, we will use Xbox Kinect camera's skeletal data mode. We will try to encode human joint information to the robotic arm which has six degrees of freedom. In order to get some understandable and logical movement, we will limit encoded while we are training the learner.
CIRAK(Apprentice) is a project that aims to create a helper robot for human workers. In this project, we are working on the understanding action of human worker and predicting the next action based on the previous ones, a motion planning algorithm that is safe and understandable for humans. Currently, we are collecting human action data from by using Xbox Kinect's skeletal data, which is giving us human joint data estimation. For robot's simulation, we are using Openrave and ROS. Currently, I am writing an Openrave controller in order to be able to control the robot from the simulation.
NAR enables you to store files in other people's computers in a secure and reliable fashion. By joining the system, you contribute the system by allocating disk space and bandwidth. Everyone gets service as much as he/she contributes to the network. This contribution/reward system is implemented in a distributed fashion. When storing files, NAR divides the file into smaller chunks and distributes them to the other nodes in the network. No one in the network has a complete access to your files; even if they did, the files are secured as explained below. All files you stored in the network is encrypted with AES. Your AES keys are encrypted with another AES key that is generated by a passphrase that you provided. This encrypted AES key is stored in a key server for later access to the network from other nodes. Everything between nodes is coordinated by coordinator servers. These servers can be managed by anyone who wants to establish them. Servers are inter-coordinated with inter-server protocols. Anyone can join the network as a coordinating server or a node as a regular user. By sending the chunks to multiple nodes, NAR creates redundancy. When nodes leave and join the system, it redistributes the files so that each file is accessible all the time. It does this operation in a smart way that will not consume the bandwidths of its users. Everything in this project is created with C++ with Boost and MySQL libraries of C++.