Neural Engineering

Transformative Technologies

Work Package: WP3
Programme: P8
Deliverable: Deliverable 8.1: "Human-like movements of a high DOF robot arm"

Deliverable due date: month 30

This document reports on the progress of work on Deliverable 8.1. All the planned tasks related to this Deliverable have been accomplished. Preliminary results have been published in the proceedings of
the international conference on numerical analysis and applied mathematics in 2014 (ICNAAM-2014), which is accessible under the following link: http://dx.doi.org/10.1063/1.4912427. A manuscript with a detailed description of the theoretical and experimental progress achieved thus far is in preparation. It will be submitted to a robotics journal.
As described in the NETT 289146 Grant Annex, the work originally planned in P8 aimed at controlling a single, high-degree of freedom robotic arm-hand system. We have extended our nonlinear constraint optimization approach to also include the more challenging problem of a synchronous dual-arm movement planner for our humanoid robot ARoS. The planner deals with 22 degrees of freedom distributed along the arms and the fingers of both hands. The approach takes into account the most recent literature concerning the kinematics features of human upper-limb reaching movements. Our results demonstrate that it is possible to plan and generate quite efficiently human-like trajectories in complex scenarios cluttered with obstacles. Consequently, real-time applications in human-robot interaction tasks become feasible.
Moreover, the efficiency of the optimization process has been even further improved for scenarios in which the robot has to perform the same movements several times. In order to reduce the time of convergence of the planner, a learning process has been designed and tested. Our results show that the computational burden of planning single-arm movements can be reduced up to 40%. Currently, we are extending the learning process to include also bimanual manipulation tasks.
To promote open access, we are developing a C++ library that will allow interested researchers from the robotics community to test the novel planner in diverse uni-manual and bimanual manipulation tasks.
The remaining part of P8, more closely related to concrete health care applications, will be reported in Deliverable 8.2.
The work also covers milestone 14.

Contributors: Gianpaolo Gulletta, SM Araújo, E Costa e Silva, MF Costa, Estela Bicho, Wolfram Erlhagen (UMinho)