Virtual workout designed for rover control systems
Prototype designs for robotic rovers are usually put through their paces in rock-strewn Mars Yard planetary test facilities. Now a comparable virtual environment has been created to assess the performance of rover control software.
This new On-Ground Autonomy Test Environment (OGATE) allows different autonomous robotic systems to be evaluated against standardised yardsticks. Its design was supported through ESA’s Networking/Partnering Initiative (NPI), supporting PhD research on space-related technology topics.
The further away that future rovers explore, the more autonomous they must become. The Moon is about as far as direct remote control is feasible – Russian controllers steered a pair of Lunokhod rovers across the surface in real time in the early 1970s.
Further afield, rovers are currently overseen through an exhaustive sequence of telecommands uplinked during each communication pass which are then implemented precisely, one at a time.
This adds up to a high workload for ground operators, along with a lot of stress between passes – what if something goes wrong?
The aim is to add much greater autonomy to future rovers, making robotic control safer and more intuitive, and allowing operators to focus on scientific issues. The preferred approach is termed ‘AI Planning and Scheduling’ (P&S) where the rover will be sent a ‘shopping list’ of activities, for which it will then generate its own action plan to accomplish them.
“The autonomous systems that will make this possible are inherently complex in nature, and some years away from flight,” explains Pablo Muñoz of the Department of Computer Engineering at the Universidad de Alcalá, creator of OGATE.
“So, in a sense, our work is intended as a first step to improve the experimental evaluation and validation of such systems.”
While various P&S control systems are similar in overall concept, individual approaches may be based on different priorities. Some, for example, might prioritise resource usage, such as conserving battery power, while others may focus on timeliness.
“Using the OGATE, we can assess how well the plan generated by a P&S system fits with the robotic execution,” adds Pablo. “We can compare how the robot achieves its goals with the minimum and maximum time range given by the planners, to provide a measure of the plan efficiency.
“We can also evaluate other aspects, such as the time taken to generate the plan in the first place, for how long the rover remains idle, the computational resources the controller requires, or how it responds to an unexpected event. We currently have 17 metrics required, and are working to extend a full set.”
For now, OGATE assumes a flat planetary surface for its case studies, but the team has performed initial simulations of the 2020 ExoMars rover in a Mars-like virtual environment, seen as a good candidate to demonstrate autonomy features.
Pablo worked with ESA’s Planetary Robotics section while developing OGATE, having spent two months at ESA’s ESTEC technical centre in Noordwijk, the Netherlands.
About the Networking/Partnering Initiative
Currently celebrating its 10th anniversary, ESA’s NPI supports work carried out by universities and research institutes on advanced technologies with potential space applications, with the aim of fostering increased interaction between ESA, European universities, research institutes and industry.
