Robotics engineers are developing algorithms to make HyQ, a quadruped robot, into a useful tool in disaster missions.
Robotics engineers in Switzerland and Italy are developing algorithms that they hope will make a dog robot robust enough for use in search and rescue missions and environmental disasters that are too dangerous to employ humans.
The HyQ quadruped robot was developed and built by Claudio Semini, of the Italian Institute of Technology (IIT) in Genoa, who completed it in 2010. A copy of the robot was built in 2012 for ETH Zurich Professor Jonas Buchli’s laboratory, part funded by the National Science Foundation. Buchli’s team is developing algorithms to increase the robot’s viability.
ETH Zurich postdoctoral researcher Thiago Boaventura told Reuters that HyQ’s actuation system makes it mechanically robust and allows it to run at speed. “One of the key characteristics of this robot is that we use an actuation system that is mechanically very robust and is strong and fast, so wheels, hydraulic activators, to drive the joints. Us humans have muscles to move the joints, we use this kind of actuations to drive and move the joints and these actuators are able to withstand impact that some kind of actuations, for example electro motors they might have some issues. This is a key characteristic of this robot here to perform and withstand impacts that are intrinsic to walking and trotting motions, for example,” said Boaventura.
Pointing to one of the actuators, he said that they were “able to control all the interaction forces that we generate with the leg. So we can essentially set the dynamical behavior of the robot using the software, so we can define the stiffness and damping parameters of the leg and all the joints using only software instead of real springs or dampers that are placed on the robot.”
Emulating the elastic elements and springiness of the robot’s joints by using software, rather than hardware, is another positive.
Boaventura said an actuator in each knee joint allows programmers to control the force applied when overcoming a variety of terrain. “Here in the knee joint we have the hydraulic actuator that’s attached to the joint through a load cell. This load cell is what allows us to control the force that we apply in this joint and consequently to the terrain, so the joints control the forces that will move the robot,” he said.
Buchli agreed that HyQ’s force control is crucial. “The force, or what we also call torque control, of this robot is a very important feature when you’re interacting with the environment and it actually is the only way to really make interaction with the environment robust, and that’s our research interest in this machine, that we have a machine that can go out there and work in environments that are not known to the robot and do these robustly, so that eventually we can use these robots to do something useful,” he said.
According to Buchli, HyQ shares many similarities with Spot, the dog robot developed by Google’s Boston Dynamics, not least its ability to withstand being kicked and pulled around when on the move. Yet he says the multinational company’s secrecy makes it difficult to understand the specifics of Spot’s functionality. Conversely the makers of HyQ have revealed many of the secrets behind their robot, encouraging others in the academic community to emulate their work.
In tests on a treadmill and outside of the laboratory, researchers at IIT, where Buchli and Boaventura previously worked, have shown for the first time that an actively controlled machine, without passive springs, can successfully walk, run and jump over obstacles and difficult terrain.
Buchli’s team at ETH Zurich is using HyQ as a research platform on so-called Optimal and Learning Control, with the aim of making it undertake useful, dynamic, tasks.
The eventual aim is for HyQ, or a future successor, to be able to be sent to conduct dirty and dangerous tasks, such as at a nuclear disaster site, instead of humans.