Last week, Team Tumbleweed’s Head of Software Engineering, Tolga Ors, attended the Global Conference on Space Exploration (GLEX 2025) in New Delhi, India. There, Tolga presented the work he and his team have been doing for the past 2 years on an algorithm to determine the Tumbleweed rovers’ location as they collect data on Mars.
Traditional Mars rovers have brought us many important insights into the Red Planet’s surface in their years of travelling across it. One major drawback of these large, wheeled rovers, however, is their speed: Curiosity, for example, has travelled less than 35 kilometres since arriving on the planet in 2012. This is one of the points our wind-driven Tumbleweed Mars rovers aim to address: Travelling at up to 18 km/h, swarms of our rovers will be able to cover large areas in a short time.
While the increased speed means that the swarm can collect data across the planet quickly, it also makes it more difficult to accurately determine the location at which a data point is collected. In contrast to Earth, Mars does not have dozens of satellites in orbit that help create a precise geolocation system. Additionally, the rovers will be rolling almost all the time. From this unique challenge, our engineers have developed a novel Location and Attitude Determination (LAD) system, suited to the Tumbleweeds’ requirements.
The LAD system combines multiple sensors. These include an Inertial Measurement Unit (IMU), which tracks the rover’s acceleration and how fast it rotates, and a stereo camera that can detect terrain features. In addition, a star tracker is used to help correct the location when the rover is stationary during Martian night.
To test their algorithm, Tolga’s team created a digital simulation environment of Mars. In this simulated environment, they were able to model the Tumbleweed’s rolling behaviour under varying wind and terrain conditions. Naturally, for a wind-driven Mars rover, realistic simulations of the wind are particularly important to model the Tumbleweeds’ rolling behaviour. While the simulation resulted in more accurate estimates for the location than in previous attempts, the error between the ground truth and the estimated location and attitude increased over time. The objective of the team is to reduce this error further by combining sensor data and using more advanced noise filters.
The work done by Tolga and his team was a good match for the theme of GLEX 2025: “Reaching New Worlds: A Space Exploration Renaissance.” Co-hosted by the Indian Space Research Organization (ISRO) and the Astronautical Society Of India (ASI), the event welcomed more than 1,400 visitors, including astronauts, scientists, policymakers, and entrepreneurs.
Sharing the results at GLEX 2025 (Paper Abstract), Tolga especially appreciated the attendees’ interest in discussing his team’s work: “The audience asked lots of great questions despite it being the last session of the event.” The plan for later this year is to compare the improved algorithm simulation results with measurements from an analog mission using a prototype Tumbleweed rover.
For further information, please contact office@teamtumbleweed.eu.