Faculty Development Competitive Research Grant Program 2018-2020
The scope of this research project is to develop bio-inspired adaptive control strategies suitable for humanoid robotics applications. By continuing the work started with a previously founded project we intend to complete the development of a light-weight middle scale humanoid robot and use it to test the effectiveness of the proposed control algorithms. In particular, our main goal is to endow the robot with the capability to learn and improve its locomotion and manipulation skills while interacting with its environment. The outcomes of this project are both scientific and technological. On the one hand we want to develop innovative control strategies that will enhance the adaptability and flexibility of humanoid robots in performing locomotion and manipulation tasks. On the other hand we want to advance the technology of humanoid robotics by achieving more lightweight and energy efficient systems. Thanks to the combination of 3D printing techniques and the usage of lightweight metals and composite materials, it will be possible to develop rigid and durable kinematics structures. By reducing the weight and the inertia of the robot's components, the system will be inherently safe in case of unexpected collisions with human beings present in the surrounding. Furthermore, due to the fact that smaller forces will be required to accelerate the robot joints, the energy consumption will be reduced and the autonomy of the system increased. We estimate that for a total robot weight of 25 kg an average electrical power of 150W is required. This will give autonomy of about two hours if a battery with a capacity of 320Wh and a weight of 3kg will be installed on-board. Some of the long-term applications we foresee for our robotic system are: monitoring and surveillance of homes and public environments, assistance of elderly people, guidance inside hospitals and museums, and caregiving to children by therapeutic games and interaction.
|Effective start/end date||3/20/18 → 12/31/20|
Recurrent neural networks