Abstract:We consider solving a cooperative multi-robot object manipulation task using reinforcement learning (RL). We propose two distributed multi-agent RL approaches: distributed approximate RL (DA-RL), where each agent applies Q-learning with individual reward functions; and game-theoretic RL (GT-RL), where the agents update their Q-values based on the Nash equilibrium of a bimatrix Q-value game. We validate the proposed approaches in the setting of cooperative object manipulation with two simulated robot arms. Although we focus on a small system of two agents in this paper, both DA-RL and GT-RL apply to general multi-agent systems, and are expected to scale well to large systems.
Abstract:While social robots are developed to provide assistance to users through social interactions, their behaviors are dominantly pre-programmed and remote-controlled. Despite the numerous robot control architectures being developed, very few offer reutilization opportunities in various therapeutic contexts. To bridge this gap, we propose a robot control architecture to be applied in different scenarios taking into account requirements from both therapeutic and robotic perspectives. As robot behaviors are kept at an abstract level and afterward mapped with the robot's morphology, the proposed architecture accommodates its applicability to a variety of social robot platforms.
Abstract:There is a lack of autonomy on traditional Robot-Assisted Therapy systems interacting with children with autism. To overcome this limitation a supervised autonomous robot controller is being built. In this paper we present a multilayer reactive system within such controller. The goal of this Reactive system is to allow the robot to appropriately react to the child's behavior creating the illusion of being alive.