A dynamic field approach to goal inference, error detection and anticipatory action selection in human-robot collaboration

Abstract

In this chapter we present results of our ongoing research on efficient and fluent human-robot collaboration that is heavily inspired by recent experimental findings about the neurocognitive mechanisms supporting joint action in humans. The robot control architecture implements the joint coordination of actions and goals as a dynamic process that integrates contextual cues, shared task knowledge and the predicted outcome of the user's motor behavior. The architecture is formalized as a coupled system of dynamic neural fields representing a distributed network of local but connected neural populations with specific functionalities. We validate the approach in a task in which a robot and a human user jointly construct a toy 'vehicle'. We show that the context-dependent mapping from action

Human-Like Movement of an Anthropomorphic Robot: Problem Revisited

“…For a definition of the grip types and on the details of the task see e.g. [2,3]. Previously in [4] we have tested the performance of three nonlinear constrained optimization solvers (IPOPT, KNITRO and SNOPT) for these four problems using a fixed discretization of time and with sligtly different constraints (see previous section).…”

“…In previous work we have presented a model for generating smooth and human-like arm and hand movements of an anthropomorphic robot, ARoS, that is inspired by the Posture-Based Motion-Planning Model (PBMP) of human reaching and grasping movements [1]. Several experiments have been made on human-robot interaction tasks (see [2,3]). However, for a fluent and efficient human-robot interaction, movements must be generated in real-time or at least in an sufficiently small interval of time.…”

A dynamic neural field architecture for a pro-active assistant robot

“…For the control of the arm-hand system we applied a global planning method in posture space that allows us to generate smooth and natural movements by integrating optimization principles obtained from experiments with humans [7]. Figure 2 presents schematically the robot cognitive control architecture which is inspired by known neuro-cognitive mechanisms underlying perception, reasoning and action in a social context (for details see [4], [3]).…”

“…In our previous work we have developed and applied a dynamic field model of action understanding and complementary action selection that implements these ideas [4], [3]. It consists of a distributed network of local pools of neurons each with specific functionality.…”

Guest Editorial: Technologies and Applications for Interactive NUX/NUI

“…Daxing Jin et al proposed an approach to generate virtual agents that can support users for NUIbased applications through human-robot interaction (HRI) learning in a virtual environment [2]. The method was implemented in a virtual environment, which improves the learning speed, efficiency and safety of the learning procedure.…”