Hannover/Aachen. In future space missions, robots will be used for increasingly complex tasks. It is envisaged that they will explore difficult-to-access terrain, such as caves and craters, on remote planets, set up infrastructure for base camps, undertake maintenance and repair work on orbiting satellites, and even clear away space junk from Earth orbit. In many of these applications, particularly ones on distant celestial bodies, remotely controlling robotic systems from Earth is simply not practicable owing to the time delays involved. That is why tomorrow’s space-mission robots will need to be autonomous actors.

Pioneering design: AI-based autonomy and multifunctional morphologies

The DFKI Robotics Innovation Center, headed by Professor Frank Kirchner, is developing autonomous robot technologies for use in space. Copiously fitted with sensors, they will be able to perceive all aspects of their surroundings. The researchers are also using AI methods and algorithms, such as machine learning, for environment perception, localization and motion planning. Together, these technologies will enable the robots not only to act and make decisions autonomously, but to learn from their own behavior as well. These technologies are vital to enabling the robots to operate in planetary and orbital missions for prolonged periods and without human intervention.

To enable robots to actually access difficult terrain of particular scientific interest on other worlds, the DFKI researchers are developing innovative, highly complex bio-inspired mobility and morphology designs. These range from multi-legged crawling and climbing robots to hybrid systems with both legs and wheels, to wheel-legged rovers with active suspension, to upright walking and climbing systems in humanoid form. Modular and reconfigurable, these systems are readily adaptable to differing conditions and tasks and are therefore capable of undertaking complex space missions on their own, in teams with other robots, or in collaboration with humans.

Intuitive teleoperation technologies and human-robot collaboration

These types of autonomous robots also need to be remote-controllable from Earth or spacecraft if the need arises, such as when the tasks in question require a high degree of flexibility. To accommodate this requirement, the DFKI researchers are developing new types of teleoperation technologies that are highly intuitive to use. One approach involves the human operator wearing an exoskeleton that provides force (haptic) feedback from the remote robot. The operator can feel when the robot encounters an obstacle and thus has the physical sensation of being part of what is happening.

It is also envisaged that robots and astronauts will one day be able to work directly alongside each other in space, such as when building and setting up infrastructure. To enable this, the DFKI research team is looking at variable autonomy – flexibly giving the robot more or less autonomy, depending on the complexity of the task at hand. If the robot gets stuck, its astronaut colleague can then intervene and teach it new behaviors. To enable smooth human-robot collaboration, the DFKI researchers are also looking at new methods of intention analysis and recognition. One potential application of this is to enable the robot to read human emotional and physical states from human physiological data and use the insights gained to optimize its action planning.

Out of the lab: Putting autonomous space robots through their paces

To ensure that these new technologies will function properly in the challenging conditions they will encounter on Mars or the moon, they are field tested here on Earth in realistic off-world scenarios in what are known as analog missions. For example, in late 2016, a team of researchers from DFKI and the University of Bremen headed to the Mars-like desert terrain of Utah to conduct a complete mission sequence designed to put their SherpaTT and Coyote III rovers to the test. The objective of the mission was to get this odd couple (the two are morphologically very distinct) to work together to form a kind of logistics chain in which they each play distinct roles in the autonomous exploration of their surroundings and taking of soil samples. Mission control was based in Bremen, Germany, and was connected to the robots in Utah via satellite link. At the heart of the mission control system was an exoskeleton, which enabled the operator to intuitively control the robots from a distance of over 8,300 km (over 5,100 mi).

In November 2017, a two-week field test took a group of DFKI researchers to Tenerife, in the Canary Islands. There they tested newly developed algorithms designed to support autonomous and semi-autonomous exploration of challenging terrain. The algorithms enabled the CREX and Asguard IV robots to explore Tenerife’s lava tubes, which are of great interest for research into space travel. Most recently – from November to December 2018 – the DFKI researchers and various European partners headed to the Moroccan desert to test a software system developed for space missions. The robotics test platform in this case was, once again, the SherpaTT hybrid walking and driving robot developed by DFKI. The new software enabled the robot to travel over 1.3 km (0.8 mi) through rugged terrain characterized by wide, open plains, steep slopes and deep canyons.

Technology transfer: Applying space technology to hostile terrestrial environments

Robotics technologies developed for space exploration have enormous transfer potential. Built for rough terrain on other worlds, they are also well suited to use in extreme and hostile environments here on Earth, such as the deep ocean or contaminated sites, for example. To achieve the necessary autonomy and capacity to act in hostile terrestrial environments, robots need to meet very similar requirements in terms of mobility, robustness and self-learning as they do in space. For example, the DFKI researchers have so far already succeeded in modifying their SherpaTT robot for a deep-sea scenario in which it is used as an autonomous under-sea rover for such tasks as sustainable resource extraction or monitoring and inspecting deep-sea plant and infrastructure. The researchers have also fitted the Coyote III micro-rover with a gas sensor so that it can be used in disaster situations to reconnoiter damaged and inaccessible buildings and sniff out gas leaks.

The researchers from the DFKI Robotics Innovation Center will be presenting their research from 1 to 5 April 2019 at the DFKI pavilion in Hall 2 (Stand C59) at HANNOVER MESSE.