Model-based off-design identification and detection software (MONSID)
Keeping NASA’s robotic explorers healthy requires smart software, especially in remote and harsh environments. An initial investment from NASA’s Small Business Innovation Research (SBIR) program has led to the development of new technology that could enable robotic exploration of distant destinations in our solar system.
Europa, an intriguing moon of Jupiter, has an icy surface and evidence of a liquid ocean below that likely contains more water than Earth’s oceans. Arthur C. Clarke acknowledged its uniqueness in his 2010 novel: Odyssey Two with the memorable passage “All these worlds are yours. Except Europe. Don’t try to land there”, as a warning to future explorers to protect this world for possible evolution of life. Scientists believe that life is possible on Europa if there is liquid water and favorable chemical elements. But you have to get closer to confirm their presence – and exploring Europe is a formidable challenge. Given the distinctive characteristics of Europe, how not to explore this fascinating destination?
To answer that call, NASA is developing concepts for robotic explorers that will land on Europa and search for signs of life by sampling the ice and possibly the ocean below. The surface of Europa is an inhospitable place, with temperatures below -260°F and bathed in strong radiation that can disrupt electronics. Europa is also so far from Earth that it would be difficult to manage a robot’s activities on an hourly or even daily basis. (It will take about 50 minutes for a signal from Earth to reach any robotic explorer that NASA deploys there.) One solution is to make our robotic explorers smarter and more autonomous, so they can detect and handle problems without human intervention. A crucial aspect of autonomy is a robot’s ability to check its hardware for problems, determine if something is not working properly, and identify the faulty component.
NASA’s Jet Propulsion Laboratory (JPL) is working with a small company called Okean Solutions, Inc. to develop this capability that could one day be used on missions in harsh space environments like Europa (Figure 1). Okean’s Model-based Off-Nominal State Identification and Detection (MONSID) software verifies the health of a system by comparing on-board measurements to simulations or models of expected behavior. MONSID checks for correct behavior, detects when something is wrong, and then identifies which piece of hardware is malfunctioning. This “model-based” approach uses the constraint suspension technique, an analytical approach developed for use on digital systems by professors Randy Davis and Howie Schrobe of the Massachusetts Institute of Technology (MIT). While working on her thesis at the University of California, Los Angeles, Lorraine Fesq extended this ability to work with analog values such as temperatures, voltages, and currents found on robotic systems. Ksenia Kolcio and Maurice Prather, Vice President and President of Okean Solutions respectively, went on to develop and mature MONSID through several efforts funded by NASA and Air Force SBIR. MONSID has demonstrated its versatility through hardware testbed deployments at JPL, the California Institute of Technology, and the Air Force Research Laboratory.
In a MONSID application, a team of engineers led by Ryan Mackey at JPL worked with Okean Solutions to model the mobility system of Athena, a JPL development rover used to test new robotic technologies (Figure 2). Modeling began when the rover was refurbished with new components, and testing with the upgraded rover yielded surprising results. The team performed tests on the Athena rover in JPL’s Mars Yard and compared data collected from the rover’s mobility system to predictions from MONSID models. Several discrepancies were observed. Initially, the team suspected the discrepancies were due to modeling errors, but further analysis revealed several new issues with the rover itself. These issues included incompatibilities between higher-level motion commands and position commands seen by motor controllers; non-nominal responses to commands, resulting in an arcing motion when straight line driving was commanded; and motor polarity mismatches between the reported steering angle and the controls, which resulted in the motors being returned to the supplier for corrections. The remaining issues were resolved by updating the motor controller software and firmware on the mobile, as well as adjusting the motor controller settings. This exercise demonstrates how the MONSID model-based approach provides smarter, more autonomous capabilities to assess hardware health and performance, even when systems are developed before deployment.
After the Athena rover was completed and checked for proper behavior, the team injected faults to simulate engine stalls and incorrect wheel controls to test MONSID’s performance under real-world conditions. For example, the team introduced a simulated bit flip that reversed the polarity of a command to one of the steering motors during a commanded drive along an arc (Figure 3). This condition caused a wheel to spin to the right instead of the left, causing that wheel to drag through the Mars Yard. The telemetry data from the mobile seemed correct despite the faulty command; in fact, the rover still achieved its propulsion goal because the other five wheels compensated for the trailing one. In the Mars Yard, engineers were able to visually see the dragged wheel, but if the rover was on the surface of another planetary body, the defect would not show up in telemetry data, making it difficult to detect. , let alone diagnose. . If left undetected, a dragging wheel can lead to serious wheel damage. Traditional limit checking approaches would be difficult to detect this type of fault because all onboard measurements remained within limits. By capturing the predicted coordination between the six wheels, MONSID was able to detect and isolate this fault immediately, unlike traditional monitoring and response methods for detecting faults. These results highlight the advantages of MONSID’s model-based approach. With MONSID, engineers can ensure not only that the right system is built before deployment on its mission in a remote location, but that hardware health can be autonomously assessed throughout the mission.
Currently, the JPL team is building MONSID models to be deployed on two new robotic systems. The first, funded by NASA PSD’s Concepts for Ocean worlds Life Detection Technology (COLDTech) program, will diagnose hardware from a robotic arm designed to pick up material from the surface of icy worlds such as Europa and Enceladus, the moon of Saturn. As part of a COLDTech task titled REASIMO (Robust, Explainable Autonomy for Scientific Icy Moon Operations), MONSID will be deployed on two robotic arm testbeds funded by NASA’s Planetary Exploration Science Technology Office (PESTO): The Ocean Worlds Autonomy Testbed for Exploration Research and Simulation (OCEANWaters) software testbed developed at NASA Ames, and the Ocean Worlds Lander Autonomy Testbed (OWLAT) hardware testbed developed at JPL.
For the second task, funded by NASA SMD through the JPL Type II Office, the MONSID team will prototype diagnostic models of spacecraft power and attitude control subsystems for the Lunar Flashlight CubeSat of the NASA, slated for launch later this year. These models will support pre-launch testing of the spacecraft, with a vision to download the models and MONSID engine during the extended mission after the main mission objectives have been achieved. What began as an SBIR investment has now led to an opportunity for MONSID to truly spread its wings, as a demonstration in space paves the way for autonomy to enable challenging and exciting new missions.
The research was conducted at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration (80NM0018D0004).
Dr. Lorraine Fesq, Jet Propulsion Laboratory, California Institute of Technology
NASA SBIR Program, JPL Research and Technology Development Program, NASA PSD COLDTech Program, Air Force SBIR Program, JPL Type II Office