|
|
|
| e-Pub |
Previous | 
Home
Section: New Results
Robotics
Mobile manipulation by humanoid robots
Participants : Pierre-Brice Wieber, Dimitar Dimitrov, Alexander Sherikov, Jory Lafaye, Camille Brasseur, Nestor Alonso Bohorquez Dorante.
This year's contributions to the field of mobile manipulation by humanoid robots have been three-fold: a lexicographic MPC approach to the decision of using optional contacts when necessary to maintain balance (and only when necessary), a robust MPC approach to online generation of dynamic walking motion on uneven ground such as stairs, and an analysis of the role of viability and capturability in collision prevention, using once again a lexicographic MPC approach.
Reactive trajectory generation
Participants : Pierre-Brice Wieber, Dimitar Dimitrov, Saed Al Homsi.
The goal of the collaboration with Adept Technologies is to generate time optimal trajectories in the presence of moving obstacles in real time. Three approaches with increasing computational complexity have been proposed and validated experimentally. The cheapest approach begins with a standard bang-bang control which is time-optimal in the absence of obstacles, and simply projected on dynamic limits imposed by collision avoidance. This leads to reasonable results where collisions are explicitly avoided, but time-optimality is lost in the process. A more complex approach introduces an MPC scheme minimizing a weighted L1-norm, which is tuned to generate a time-optimal behavior in the absence of obstacles. In the presence of obstacles, time-optimality is once again lost, however, results are much improved with respect to the previous approach. The final, and most complex approach, considers time-optimality as a lexicographic objective: a lexicographic MPC scheme is proposed, which achieves time-optimality in the presence of obstacles, with reasonable online computation time. This work has been submitted to ICRA 2016.