研究方向的特化:机器人学的研究方向
协作机器人Collaborative & Shared Autonomy
- Full-autonomy challenges due to:
- close interaction with multiple objects
- noisy sensing with ambiguity
- hard to model real-world dynamics
- need for guarantees for safe operations
- highly un-constrained environment
- un-modelled user intentions
Optimal Feedback Control
- Choices in Planning with OFC
- Dynamics Transition Models How to acquire/learn dynamics or incrementally improve models?
- Representation What is a suitable representation of desired movement (trajectories, goal)
- Choice of cost function How to design a cost function for desired movement?
- Exploitation of natural dynamics How to optimize spatiotemporal parameters for energy efficiency? Can we exploit variable impedance?
形态计算Morphological Computation
- reconfigurable
- modular
- scalable (size and number)
from single robot designed for single task, to multitasking
- 通过形态学简化或替代控制理论
replacing/simplifying control by morphologies
- 控制理论与物理模型 PD-control vs. Spring-damper system
- 主动控制与被动适应 Fully-actucated bipedal locomotion vs. Passive dynamic walking
- 手眼协调与通过设计扩展夹爪适应性 Hand-eye coordination vs. Universal gripper(如A Positive Pressure Universal Gripper Based on the Jamming of Granular Material)
What‘s next in Morphological computation
- Theoretical foundation
- Computation for embodied agents
- Energetics and efficiency of computation
- Computation across spatio-temporal scales
- Programmability, repoducibility/accuracy
- Technological Innovations
- Material-level sensory-motor components
- 3D-printing, robot-building-robot
- Material-level programmability
- Environment-inclusive programmability
- “Common currency/languages” between brain and body
- Emergence of “digital computation”
- Integrate and integrate Neural Networks