Valkyrie stands more than six feet tall, weighs 286 pounds, and has an 80 inch wingspan. “It feels human-like, you can look her in the eyes,” says Reg Berka, deputy project manager of Valkyrie, the NASA built robot destined for the red planet.
Drawing on a heritage of building humanoid robots, the JSC team had to overcome many challenges throughout the process – primarily, dealing with gravity. “There are things that we as NASA engineers are familiar with, and there are things that we are not as familiar with,” explains Berka. “One of these big things we don’t’ know well is gravity.”
But the processes is not over. “It’s half-time,” says Berka, as the team continues to develop Valkyrie, preparing for the DRC Finals down the road. “We knew it was going to take until the finals until this complex machine was going to be able to really preform.”
To begin the design of this complex machine, the team started with paper sketches, designing a number of different configurations. However, because of the funding constraints and time restrictions (and a government furlough mid-project) the team didn’t prototype.
“We built the robot and tested it. There wasn’t time for anything else,” explains Berka. “Certainly, I would have built two robots.”
Instead, Valkyrie was built up joint by joint. “We used single joint test beds to characterize and program the controllers and to ensure that we were going to get the response out of the joints that we needed,” says Berka.
Working in parallel with the hardware team, the software was developed separately from the beginning, not phased with the physical robot until the end.
“In order for that to happen, it was a testament to the automation tools, both hardware and software,” says Berka.
The tools included a product called “Gazebo,” a DARPA specified program managed by the Open Source Robotics Foundation. A virtual, simulation-system-based program, Gazebo was enhanced after early test bed results to improve its capabilities.
These capabilities, like how to take a step forward and how to open a door, are programmed into the robot, but high level commands, like determining what a door is, are relayed by the robot’s human counterpart.
“If you stand a robot in a room it’s very hard to say ‘find the door,’ because robots aren’t as perceptive as people. Even if you told it what a door was, if it wasn’t exactly the same door that you programed it for, it would miss it,” explains Berka. “There is always this weakness of perception. A robot can sense its environment, but knowing a door from a cabinet is hard thing for a robot.”
In order to navigate its environment, a series of video cameras, laser scanners, and ultrasonics provide the team controlling the robot with a detailed view of the surroundings. “Over 100 different sensors are fused together to build an image,” adds Berka.
Working at a higher level, three computers inside Valkyrie’s chest serve to determine what the robot should be doing, and then command the joints to carry out those movements.
The “brains” of the robot, the computers are controlled by a human counterpart behind the scenes.
“It’s really a pretty conventional system in that it’s comprised structurally of aluminum,” says Berka. The robot’s secondary structure is made entirely from 3D printed plastic, including mounts for the video cameras, and support brackets for the electronics; yet, unlike other robots in the competition Valkyrie is electric, not hydraulic.
“Hydraulics take a lot of power and tends to leak fluid. Since we build for space and spacecraft, we would never use hydraulics,” explains Berka. Valkyrie was also one of the only robots with the ability to work untethered, and can operate for nearly an hour on a single charge.
For future versions of the robot the team is exploring the use of advanced composite materials to reduce the robots weight, while improving its strength.
Designed for Mars
Even though the team is not happy with their performance in the trials, their sites are set on much higher aspirations – trying to one day support a mission scenario to Mars.
Valkyrie tied for last place (with two other teams), scoring zero points after a bug was inadvertently introduced into the software on the day of the challenge, rendering it unable to walk. The problem was fixed by the end of the day, but not in time for the competition.
“The DARPA program manager made, what I consider to be a rather bold prediction, saying that he thinks the NASA robot is the only robot really capable of preforming at the finals at they have it envisioned,” says Berka. “Out of the box, we are the only team that has the dexterity and manipulation capabilities, and the untethered feature.”
Berka acknowledges that the team is working to build a technology using public money, so his main goal is to have a robot that will allow them to build an outpost on the moon or Mars in advance of humans being there. “We have to make sure everything is working so when we bring humans to Mars they have a place to go,” he adds.
While many are looking at 2030 as an achievable date for such a mission, Berka believes that Valkyrie will be ready to support the undertaking long before that. Possibly in the next few years. “Our tech is well enough along that as soon as someone decides to go to Mars, I think Valkyrie will be a part of it.”
Until then, Valkyrie has one more stop at the DARPA robotics challenges finals, where the team hope to prove superior with the most sophisticated robot in the field.