Contents
Vol 5, Issue 49
Focus
- In the soft grip of nature
Biological grippers can inspire the development of a new class of versatile soft grippers in agrorobotics and beyond.
- Soft sensors that can feel it all
Soft materials and machine learning combine to enable a sensor that distinguishes bending, stretching, and compression.
- Robots learn to identify objects by feeling
Multimodal tactile sensors help robot hands accurately identify grasped objects by measuring thermal properties and contact loads.
- Neuroengineering challenges of fusing robotics and neuroscience
Advances in neuroscience are inspiring developments in robotics and vice versa.
- Reflections on the future of swarm robotics
Swarm robotics will tackle real-world applications by leveraging automatic design, heterogeneity, and hierarchical self-organization.
- Magnetic movement under the spotlight
Composite hydrogel robots can achieve programmable locomotion using light and magnetic fields.
Research Articles
- Heterogeneous sensing in a multifunctional soft sensor for human-robot interfaces
Heterogeneous sensing mechanisms enabled a multifunctional compact soft sensor to decouple multiple deformation modes.
- Skin-inspired quadruple tactile sensors integrated on a robot hand enable object recognition
A robot hand integrated with quadruple tactile sensors can recognize a diverse set of objects for garbage sorting.
- Multi-expert learning of adaptive legged locomotion
A multi-expert learning architecture generates adaptive behaviors for the versatile locomotion of quadruped robots.
- Fast and programmable locomotion of hydrogel-metal hybrids under light and magnetic fields
Reconfiguring and programming the action of a magnetically driven robot using light.
About The Cover

ONLINE COVER You Can Teach a Robot Dog New Tricks. Learning to adapt in unknown situations is key for robots to operate effectively in the wild. Inspired by the biomechanical control of muscular systems, Yang et al. developed a framework based on machine learning called multi-expert learning architecture to teach multiskill locomotion to a quadruped robot. Their controller consists of eight deep neural networks that represent expert skills that are then combined using a gated neural network to achieve complex locomotion such as coherent trotting, steering, and fall recovery. This month's cover is a photograph of Jueying, a quadruped robot, demonstrating agile motion. [IMAGE CREDIT: KAI YUAN AND CHRISTOPHER MCGREAVY]