by Legged robots that mimic the body structure and movements of animals have the potential to complete missions in different environments. However, these robots must be able to navigate various terrains, such as soil, sand, and grass, without losing balance or getting stuck. Researchers from the Norwegian University of Science and Technology (NTNU) and the Indian Institute of Technology Bombay have developed a new artificial paw with sensing capabilities to enhance the mobility of legged robots.
The researchers, in a paper posted on the preprint server arXiv, introduced a sensorized paw that can identify different terrains and their properties by estimating the force applied to its surface from the ground. Their previous research for the DARPA Subterranean Challenge highlighted the importance of robust response to challenging terrains. In collaboration with ETH Zurich, the team used the legged robot ANYmal to participate in the competition. Recognizing the limitations of existing technology, they concluded that enhancing a legged robot’s perception through sensorized paws could improve locomotion control.
Moving legged robots on uneven and complex terrains has been a significant challenge. Previous studies have revealed that difficult terrains can restrict the movements of these robots and hinder their ability to sense their surroundings effectively. To address this, researchers have been developing computational methods to recognize different terrains and modify legged robot movements accordingly. However, these approaches often rely on integrated sensors like LiDAR and cameras, which provide limited information about the terrain.
The team sought to develop a system that could gather detailed real-time information about the terrain. They created TRACEPaw, an artificial paw with a silicone-based hemispherical point end-effector. This end-effector deforms in response to contact forces and has an embedded micro camera and microphone to estimate force vectors and recognize various terrain types, including snow, sand, and gravel. The system collects sensory data from the terrain below it, which is then analyzed by a computer vision model trained through supervised learning to predict terrain information and estimate contact force.
One advantage of this sensing system is that it was created using off-the-shelf electronic components, making it easily and affordably scalable. The researchers conducted a series of experiments to evaluate the system’s performance, showing that TRACEPaw significantly enhances the mobility and adaptability of legged robots. It enables them to recognize and respond to different terrains, preventing incidents like slipping or stumbling.
In the future, the artificial paw could be deployed in real-world settings for search and rescue or exploration missions. The team plans to further improve the system by training its algorithm on more data to refine its force estimation and soil classification capabilities. They also aim to enhance the system’s understanding of the environment by incorporating data from an on-board IMU, providing insights into terrain slope and force direction.
Enhancing the mobility of legged robots on various terrains is a crucial step towards their practical application in different scenarios. By developing a sensorized paw, researchers have made significant progress in improving the adaptability and responsiveness of legged robots, making them more reliable and efficient in navigating dynamic environments.
Note:
1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it
