Amazon and University of Michigan give robots a sense of touch

Researchers from Amazon and the University of Michigan have developed HydroShear, a simulator that accurately models tactile forces, enabling robots to learn dexterous, contact-rich manipulation policies entirely in simulation with a 93 percent average success rate. HydroShear's key innovation is the addition of path-dependent force tracking to hydroelastic contact models, allowing for accurate tracking of forces accumulating over a soft sensor membrane during physical interactions. This approach enables the simulation of subtle tactile forces and shear feedback, bridging the tactile reality gap. The practical implication for engineers building AI systems is the potential to develop more advanced robotic manipulation capabilities.

⚡ Key Takeaways

• HydroShear achieves a 93 percent average success rate across four challenging tasks.
• The simulator uses path-dependent force tracking to accurately model tactile forces.
• HydroShear is GPU parallelizable, enabling efficient large-scale policy training.
• The simulator handles full 3-D motion, including tilting and rolling, essential for dexterous manipulation.
• Calibration is done by collecting controlled real-world data with a robot arm and vision-based GelSight Mini tactile sensors.

The development of HydroShear has significant implications for engineers building AI systems, as it enables the creation of more advanced robotic manipulation capabilities, which can be applied to various real-world applications, such as warehouse automation and surgical assistance. This technology can improve the efficiency and accuracy of robotic tasks, leading to increased productivity and redu

✅ Practical Steps

1. Implement HydroShear in robotic simulation environments to improve manipulation capabilities.
2. Calibrate HydroShear using controlled real-world data with a robot arm and vision-based tactile sensors.
3. Utilize HydroShear's GPU parallelizability to enable efficient large-scale policy training.

Want the full story? Read the original article.