Development of a Basic Educational Kit for Robotic System with Deep Neural Networks

Abstract

In many robotics studies, deep neural networks (DNNs) are being actively studied due to their good performance. However, existing robotic techniques and DNNs have not been systematically integrated, and packages for beginners are yet to be developed. In this study, we proposed a basic educational kit for robotic system development with DNNs. Our goal was to educate beginners in both robotics and machine learning, especially the use of DNNs. Initially, we required the kit to (1) be easy to understand, (2) employ experience-based learning, and (3) be applicable in many areas. To clarify the learning objectives and important parts of the basic educational kit, we analyzed the research and development (R&D) of DNNs and divided the process into three steps of data collection (DC), machine learning (ML), and task execution (TE). These steps were configured under a hierarchical system flow with the ability to be executed individually at the development stage. To evaluate the practicality of the proposed system flow, we implemented it for a physical robotic grasping system using robotics middleware. We also demonstrated that the proposed system can be effectively applied to other hardware, sensor inputs, and robot tasks.

Composition of Robot Motions based on the Concept of Deep Predictive Learning

“…The proposed framework comprises lightweight models, and the computational time and cost required for motion generation are low. Moreover, implementing each of the previous functions as components makes it possible to easily reuse the implemented system when tasks or robot hardware changes, or devices are added [15].…”

Self-Supervised Learning of Visual Servoing for Low-Rigidity Robots Considering Temporal Body Changes

“…Next, the steps from 3 to 8 are performed in exactly the same manner except for the opening and closing of the gripper. The robot reaches its hand out to the exact same place as in 5 , makes the pregrasp posture at 10 , grasps the object at 11 , and returns to the initial posture. By repeating this procedure many times, it is possible for the robot to autonomously collect data.…”

“…This is based on the fact that while it is difficult for the robot to directly grasp an object by visual recognition, it can grasp an object by reaching out to the exact same place if the object had been placed by itself. There is a similar data collection method [11], but it focuses only on automatic data collection for a rigid robot, and its goal is different from this study, which utilizes the motion reproducibility of a low-rigidity robot. Although our data collection method is limited to pick-and-place tasks, it is a basic motion common to various tasks, and we believe that it would be useful for cost effective robot that cannot move accurately due to low rigidity to learn such tasks autonomously.…”