Au-Spot, an AI robot dog, trains to explore caves on Mars

Scientists are equipping four-legged animal-mimicking robots with artificial intelligence (AI) and a suite of sensory equipment to help bots navigate treacherous terrain and underground caverns on the Red Planet.

NASA and JPL-Caltech are creating Martian dogs that can maneuver like iconic wheeled ATVs such as Spirit, Opportunity, Curiosity, and the recently released Perseverance, and even do what they can’t. The new robots’ agility and stability are combined with sensors that allow them to avoid obstacles, choose between multiple paths and build virtual maps of buried tunnels and caves for operators at their home base, the scientists said.

Traditional rovers are limited to most flat surfaces, but many scientifically interesting Mars regions are only reachable on very rugged terrain or underground. Walking robotic dogs are well suited for such tasks – even if they fall, they can get up again.

The Mars Dog will also be about 12 times lighter than modern rovers and will be able to travel much faster, reaching a normal speed of 5 km / h during testing. Simultaneously, the Curiosity rover moves along the surface of Mars at a speed of about 0.14 km / h.

According to NASA, caves on Mars could provide shelter for future human colonies, providing natural protection against deadly ultraviolet radiation, extreme cold, and violent dust storms that can last for weeks and are sometimes large enough to be seen with telescopes on Earth. Caves can also serve as a source of evidence for life from the distant past of Mars or even serve as home to organisms that live deep underground. Robots on legs can walk around rocks, descend into caves, choose a path, and collect measurements and build a map of what they “see.””

Mars’s autonomous robot, dubbed Au-Spot, is a modified version of Spot, a four-legged mechanical explorer created by the robotics company Boston Dynamics. More than 60 scientists and engineers on the Collaborative SubTerintage Autonomous Resilient Robots (CoSTAR) team have equipped the Au-Spot with networked sensors and software to help it scan, navigate and map its environment safely and autonomously.

The Au-Spot processes input from lidars (remote sensing using laser pulses), visual, thermal, and motion sensors to create 3D maps. Mars Dog also uses AI to learn which structures to avoid and identify objects of scientific interest. Simultaneously, the communications module allows the robot to transmit data to the surface while exploring the underground.

CoSTAR team members test Au-Spot on various obstacle courses, overcoming difficulties in tunnels and corridors, upstairs and ramps, and outdoors, simulating Martian landscapes, such as in lava tubes in Northern California. These demos show that untethered robots can navigate boulders and map deep caves.

Google News button
Tags: