This article is part of our exclusive IEEE Journal Review Series in partnership with IEEEX.
Four-legged robots have made great strides in their ability to safely navigate complex terrain. In a recent study, researchers from Hong Kong developed a new cartographic model for four-legged robots this allows them to independently crawl under and over significant obstacles to reach their desired end point.
Researchers describe a robot that uses multi-layer mapping to understand its environment, in study published August 4th in IEEE Letters on Robotics and Automation.
Mobile bipeds and quadrupeds animals can adapt to different landscapes. Robots that can navigate similarly complex environments are attractive because they can perform missions that would be dangerous for humans to complete, such as monitoring and assessing unstable debris sites after earthquake. But ensuring that robots can effectively navigate complex environments as well as overcome different types of obstacles (such as jumping over gaps or climbing tall objects) is a challenge.
Advanced terrain mapping for robots
Peng Lu, Associate Professor at the University of Hong Kong, along with his graduate student Yeke Chen and other team members, sought to create a robot that could overcome these obstacles. To help their robot perceive its environment in detail, they developed a model that creates a multi-layered elevation map based on data from the robot's sensors. The map can display the characteristics of a wide range of landscapes using lidar data.
The team used simulation to teach the robot to recognize different terrains it might encounter in the real world. This includes very difficult terrain to navigate, such as a chasm that he must jump over, or crawling under obstacles with a protruding ledge. If the robot lacks sensor data, it can compensate to some extent by using estimates based on its training data.
“By learning various modeling skills and knowledge distillation, the robot can switch between different skills to overcome different obstacles,” says Lu.
In their study, the scientists tested their mapping technique using Unitry Go1 robot in a series of indoor and outdoor experiments where it had to independently crawl, climb or jump to overcome obstacles.
“The results show that a multi-layer height map can effectively map a variety of complex terrain, allowing the robot to easily understand its environment,” Lu says, noting that the robot was also able to autonomously switch between crawling, jumping and climbing modes as needed.
He adds that the robot unintentionally has path-planning abilities even though it was not programmed to do so. For example, when a robot encounters obstacles that are too high to overcome, it bypasses the obstacle and hence finds its way through the environment on its own through trial and error.
Lu notes that while the robot's key advantage is its ability to navigate varied and complex terrain, it can only rely on data it has already been trained on and cannot learn directly from real-world data.
Lu says his team could commercialize the robot to perform inspections on construction sites, for example, and plans to use real-world data to further improve the robot's ability to handle any type of terrain.
Articles from your site
Related articles on the Internet
