Abstract
This paper contributes to the literature on energy efficient gaits on unknown terrains for humanoid robots, the locomotion system of which has anthropomorphic characteristics. In this work, we firstly present an energy efficient gait for humanoid robots. The main feature of the new gait is the absence of an area of support. The stiffness-free foot can rotate freely around the ankle joint. This feature makes the gait suited for uneven terrains. We then present a group of neural network controllers to regulate the sagittal and lateral motion of the robot's gait in the presence of an unknown terrain. The resulting gait evaluated on an Aldebaran Nao robot, (1) reduces the energy consumption by 41% on a flat ground compared to the conventional Aldebaran gait, (2) can handle small disruptions caused by an uneven terrain, and (3) looks more like a human gait. A video showing the gait in the simulator is available. (c) 2017 Elsevier B.V. All rights reserved.
Original language | English |
---|---|
Pages (from-to) | 111 - 122 |
Number of pages | 12 |
Journal | Neurocomputing |
Volume | 280 |
DOIs | |
Publication status | Published - 6 Mar 2018 |
Keywords
- Bipedal walk control, Neural network controller, Uneven terrain
- Bipedal walk control
- ROBOT
- Neural network controller
- Uneven terrain
- LOCOMOTION
- FOOT