Interactive Imitation Learning of Bimanual Movement Primitives

Abstract--Performing bimanual tasks with dual robotic setups can drastically increase the impact on industrial and daily life applications. However, performing a bimanual task brings many challenges, like synchronization and coordination of the singlearm policies. This article proposes the Safe, Interactive Movement Primitives Learning (SIMPLe) algorithm, to teach and correct single or dual arm impedance policies directly from human kinesthetic demonstrations. Moreover, it proposes a novel graph encoding of the policy based on Gaussian Process Regression (GPR) where the single-arm motion is guaranteed to converge close to the trajectory and then towards the demonstrated goal. Factory assembly, logistics, and household applications of bimanual robots have been known for decades [7], [8]. Modern society is faced with the lack of workforce in various However, the increased number of Degrees of Freedom repetitive jobs like re-shelving products in supermarkets (DoFs) (the curse of dimensionality) implies an increased or handling heavy luggage in airports. Robots appear to be teaching complexity and the necessity of skilled human teachers the most promising solution to mitigate the negative effects of who knows how to interface with the bimanual robotic the declining workforce and perform these various complex platform. To work in variable and unstructured environments, In this paper we contribute with the Safe Interactive Movement robots must be dexterous and intelligent to quickly learn the Primitive Learning (SIMPLe) algorithm and propose: job while interacting safely with other robots, objects, and humans.