GI-NNet \& RGI-NNet: Development of Robotic Grasp Pose Models, Trainable with Large as well as Limited Labelled Training Datasets, under supervised and semi supervised paradigms

Grasp manipulations by COBOTS (COllaborative roBOTS) are supposed to be functioning the same way we grasp objects, even complex objects. However, for robots executing an intelligent and optimal grasp efficiently, the way we grasp objects, is quite challenging. The reason being that we acquire this skill by spending a lot of time in our childhood trying and failing to pick things up, and learning from our mistakes. For robots we can’t wait through the equivalent of an entire robotic childhood. To streamline the process, in the present investigation we propose to use deep learning techniques to help robot learn quickly how to generate and execute appropriate grasps. More specifically, we develop two models. One a Generative Inception Neural Network (GI-NNet) model, capable of generating antipodal robotic grasps on seen as well as unseen objects. It is trained on Cornell Grasping Dataset (CGD) and performed excellently by attaining 98.87% grasp pose accuracy by detecting the same from the RGB-Depth (RGB-D) images for regular as well as irregular shaped objects while it requires only one third of the network trainable parameters as compared to the State-Of-The-Art (SOTA) approach [1]. However, to attain this level of performance the model requires the entire 90% of the available labelled data of CGD keeping only remaining 10% labelled data for testing, making it vulnerable for the poor generalization. Also, as we all know, getting sufficient and quality labelled dataset is becoming increasingly difficult keeping in pace with the requirement of gigantic networks. To address these issues, we subsequently propose to attach our model as decoder with a semi-supervised learning based architecture known as Vector Quantized Variational Auto Encoder (VQVAE), which we make to work efficiently when we train it with available labelled dataset as well as unlabelled data [2]. Our proposed GI-NNet integrated VQVAE model, which we name as Representation based GI-NNet (RGI-NNet), has been trained with various splits of label data on CGD with as minimum as 10% labelled dataset together with latent embedding generated from VQVAE up to 50% labelled data with latent embedding obtained from VQVAE. The performance level, in terms of grasp pose accuracy of RGINNet, varies between 92.13% to 95.6% which is far better than many other existing SOTA models trained with only labelled dataset. For the performance verification of both GI-NNet and RGI-NNet models, we use Anukul (Baxter) hardware cobot and it is observed that both the proposed models performed significantly better in real time table top grasp executions. The logical details of our proposed models together with an in depth analyses have been presented in the paper. Keywords—Intelligent robot grasping, Generative Inception 1Student at Center of Intelligent Robotics, Indian Institute of Information Technology, Allahabad, Prayagraj, India-211015 priyashuklalko@gmail.com 2Winter intern at Center of Intelligent Robotics, Indian Institute of Information Technology, Allahabad, Prayagraj, India-211015 3Professor at Center of Intelligent Robotics, Indian Institute of Information Technology, Allahabad, Prayagraj, India-211015 gcnandi@iiita.ac.in Neural Network, Vector Quantized Variational Auto-Encoder, Representation based Generative Inception Neural Network.