Tag: localization

NASA’s Perseverance doesn’t need GNSS to find itself on Mars

News from NASA’s Jet Propulsion Laboratory

A new technology called Mars Global Localization lets Perseverance determine precisely where it is, without human help.

Imagine you’re all alone, driving along in a rocky, unforgiving desert with no roads, no map, no GPS, and no more than one phone call a day for someone to inform you exactly where you are. That’s what NASA’s Perseverance rover has been experiencing since landing on Mars five years ago. Though it carries time-tested tools for determining its general location, the rover has needed operators on Earth to tell it precisely where it is — until now.

A new technology developed at NASA’s Jet Propulsion Laboratory in Southern California enables Perseverance to figure out its whereabouts without calling humans for help. Dubbed Mars Global Localization, the technology features an algorithm that rapidly compares panoramic images from the rover’s navigation cameras with onboard orbital terrain maps.

Running on a powerful processor that Perseverance originally used to communicate with the Ingenuity Mars Helicopter, the algorithm takes about two minutes to pinpoint the rover’s location within some 10 inches (25 centimeters). Mars Global Localization was first used successfully in regular mission operations on Feb. 2, then again Feb. 16.

“This is kind of like giving the rover GPS. Now it can determine its own location on Mars,” said JPL’s Vandi Verma, chief engineer of robotics operations for the mission. “It means the rover will be able to drive for much longer distances autonomously, so we’ll explore more of the planet and get more science. And it could be used by almost any other rover traveling fast and far.”

This panorama from Perseverance is composed of five stereo pairs of navigation camera images that the rover matched to orbital imagery in order to pinpoint its position on Feb. 2, 2026, using a technology called Mars Global Localization. (Credit: NASA/JPL-Caltech)

The upgrade is especially valuable given how well Perseverance’s auto-navigation self-driving system has been working. Enabling the rover to re-plan its path around obstacles en route to a preestablished destination, AutoNav has proved so capable that the distance Perseverance can drive without instructions from Earth is largely limited by the rover’s uncertainty about its whereabouts. Now that it can stop and determine its exact location, Perseverance can be commanded to drive to potentially unlimited distances without calling home.

Implementation of Mars Global Localization comes on the heels of another innovation from the Perseverance team: the first use of generative artificial intelligence to help plan a drive route by selecting waypoints for the rover, which are normally chosen by human rover operators. Both technologies enable Perseverance to travel farther and faster while minimizing team workload.

Beyond visual odometry

Unlike on Earth, there is no network of GPS satellites in deep space to locate spacecraft on planetary surfaces. So missions — whether robotic or crewed — must come up with other ways to determine their location.

The Mars Global Localization algorithm runs on a fast commercial processor in the Helicopter Base Station — the upper, gold-colored box that was integrated into NASA’s Perseverance rover in a clean room. Perseverance used the base station to communicate with the now-retired Ingenuity Mars Helicopter. (Credit: NASA/JPL-Caltech)

As with NASA’s previous Mars rovers, Perseverance tracks its position using what’s called visual odometry, analyzing geologic features in camera images taken every few feet while accounting for wheel slippage. But as tiny errors in the process add up over the course of each drive, the rover becomes increasingly unsure about its exact location. On long drives, the rover’s sense of its position can be off by more than 100 feet (up to 35 meters). Believing it may be too close to hazardous terrain, Perseverance may prematurely end its drive and wait for instructions from Earth.

“Humans have to tell it, ‘You’re not lost, you’re safe. Keep going,’” Verma said. “We knew if we addressed this problem, the rover could travel much farther every day.”

After each drive comes to a halt, the rover sends a 360-degree panorama to Earth, where mapping experts match the imagery with shots from NASA’s Mars Reconnaissance Orbiter (MRO). The team then sends the rover its location and instructions for its next drive. That process can take a day or more, but with Mars Global Localization, the rover is able to compare the images itself, determine its location, and roll ahead on its preplanned route.

“We’ve given the rover a new ability,” said Jeremy Nash, a JPL robotics engineer who led the team working on the project under Verma. “This has been an open problem in robotics research for decades, and it’s been super exciting to deploy this solution in space for the first time.”

The small team began working in 2023, testing the accuracy of the algorithm they’d developed using data from 264 previous rover stops. The algorithm compared rover panoramic photos to MRO imagery and correctly pinpointed the rover’s location for every single stop.

How Ingenuity helped

Key to Mars Global Localization is the rover’s Helicopter Base Station (HBS), which Perseverance used to communicate with the now-retired Ingenuity Mars Helicopter. Equipped with a commercial processor that powered many consumer smartphones in the mid-2010s, the HBS runs more than 100 times faster than the rover’s two main computers, which, built to survive the radiation-heavy Martian environment, are based on hardware introduced in 1997.

As a technology demonstration designed to test capabilities, the Ingenuity mission was able to risk employing more powerful commercial chips in the HBS and the helicopter even though they hadn’t been proven in space. It paid off: Expected to fly no more than five times, the rotorcraft completed 72 flights.

The power of the HBS processor inspired Verma to look for ways the Perseverance mission might harness it. “It’s almost like a gift. Ingenuity blazed the trail, proving we could use commercial processors on Mars,” Verma said.

Tapping into the HBS computer has had its challenges. To address reliability, the team developed a “sanity check”: The algorithm runs on the HBS multiple times before one of the rover’s main computers checks to ensure the results match. During testing, the team repeatedly found the rover’s position was off by 1 millimeter. They discovered damage to about 25 bits — a minuscule fraction of the processor’s 1 gigabyte of memory — and developed a solution to isolate those bits while the algorithm runs.

Alongside the broader Mars Global Localization process, the team’s sanity check and memory solutions are expected to find new uses as faster commercial processors are employed in future missions. In the meantime, the team has already turned their sights to the Moon, where difficult lighting conditions and long, cold lunar nights make knowing exactly where spacecraft are located all the more critical.

More about Perseverance

NASA’s Jet Propulsion Laboratory, which is managed for the agency by Caltech, built and manages operations of the Perseverance rover on behalf of NASA’s Science Mission Directorate in Washington, as part of NASA’s Mars Exploration Program portfolio. Learn more about Perseverance.

March 3, 2026
Robosat partners seek improved localization of autonomous machines
Researchers from Finland, Switzerland, Spain and Romania gathered at Tampere University in Finland for a workshop this month within the Robosat project focusing on localization of autonomous machines.

Workshop participants discussed and demonstrated novel technical solutions to improve localization, particularly of autonomous machines operating in challenging and unconstrained environments, such as forests and mountainous regions.

The Robosat project aims to change how autonomous robots navigate in the wild by integrating multi-sensor and multi-GIS data. During the Tampere workshop, partners from Tampere University (Finland), ETH Zürich (Switzerland), Universitat de València (Spain) and CITST (Romania) discussed strategies for sharing data, identifying relevant GIS and GNSS datasets, and leveraging AI for autonomous labeling of large-scale data.

Key topics included the integration of multi-sensor and multi-GIS data to enhance positioning accuracy, planning piloting tests with ETH’s ANYmal robot and TAU’s new I/Q GNSS grabber device, and discussing methods for AI-driven data labeling for massive datasets collected during field trials.

The Tampere University project team includes Elena Simona Lohan and Jari Nurmi as supervisors and Ph.D. students Yelyzaveta Pervysheva and Muhammad Safi.

The Robosat efforts supports applications in robotics, environmental monitoring, and industrial automation. By combining expertise across Europe, Robosat intends to pave the way for smarter, safer and more efficient autonomous systems.

It also aims to provide new open-access rich datasets to the research community. A first dataset enabling multimodal classification studies has already been published on Zenodo as a collaborative work between Tampere University and CITST teams.

The Robosat project

Autonomous robot navigation in the wild using satellite-based 3D geographical information (ROBOSAT) aims to provide a scalable MultiGIS high-quality data collection platform through the use of a quadrupedal robot that can autonomously perform long-distance missions in challenging environments, such as Alpes mountains or Finnish forests.

Consortium organizations are comprised of three universities and one SME:
- Tampere University, Finland. Expertise: GNSS, wireless positioning, sensing, and communications, RF Fingerprinting and interference mitigation. Coordinator: Elena Simona Lohan
- ETH, Switzerland. Expertise: automation, mapping, control theory, and legged-robot research. PI: Marco Hutter
- Universitat de Valencia, Spain. Expertise: computer science, database management, machine learning. PI: Joaquin Torres Sospedra
- CITST, Romania. Expertise: machine learning/artificial intelligence, robotics, exploitation. PI: Irina Mocanu.
December 15, 2025
Swift Navigation, SolarCleano: cleaning robots keep solar power running

A SolarCleano F1A robot tackles a tough cleaning challenge on a solar farm in Saudi Arabia. (Photo: SolarCleano)

SolarCleano, based in Garnich, Luxembourg, makes robots that clean large solar panel installations using GNSS receivers and corrections from Swift Navigation. We asked Christophe Timmermans, SolarCleano’s managing director, a few questions about its technology.

How often do solar panels need to be cleaned?

For decades, it was believed that solar panels did not need to be cleaned due to their angle to the ground and rain. Nowadays, however, the cleaning of solar panels is widely accepted as necessary to optimize a plant’s return on investment (ROI).

How much time per sq. meter do your machines take to clean solar panels?

To provide the fastest possible ROI to our customers, we developed a range of robots to best address the needs of various solar plant layouts. A large utility-scale project with high level of soiling losses in a desert environment will need a very fast and reactive cleaning solution such as our SolarBridge B1, which can clean 24/7/365 fully autonomously. The most suitable solution for a farm rooftop in Germany that needs to be cleaned three to four times a year might be our F1 model, which can clean the equivalent of up to two soccer fields a day. It is designed for rooftops, floating panels and mid-size plants up to 50 MW. While the speed of cleaning is a very important variable, the quality of cleaning is often considered as the driver to performance, which is why we propose different types of brushes depending on the soiling types. Plus, the robot speed can be modified according to the soiling level.

Why do robots need GNSS receivers to clean solar panels?

Moving on inclined, wet glass surfaces makes odometry unreliable because robots might occasionally slip. Therefore, GNSS is the most reliable way to continuously monitor their exact position. Our robots also need path planning because they cannot operate randomly like lawn mowers. Safety is obviously a major concern; we need a very high localization accuracy to ensure that robots don’t fall off the panels. Finally, the largest solar plants are developed in dry, remote locations with high irradiation such as the Sahara, Atacama and Australian deserts. GNSS allows us to have very accurate localization even in those remote areas. In addition, this solution can easily be installed on already-existing solar plants with little capital expenditure.

What spatial accuracy requirements do the robots have for this task?

Safety is our absolute priority. Therefore, our robots need an accuracy of less than 3 cm. They also need to be aware, in real time, of changes in their surroundings, such as maintenance teams, animals and uneven ground.

On large solar farms, GNSS receivers always have a clear line of sight to the satellites and do not suffer from multipath. So, what are the key technical challenges?

Our robots have the additional advantage that they do not need to drive very fast. However, we need to manage fleets of robots on the other side of the world in regions difficult to access and with harsh weather conditions, such as very high or low temperatures and the accumulation of dust behind panels due to air vortices. We need to be able to perform remote maintenance and solve any issue from our control center in Luxembourg. These challenges make our robots increasingly robust. With a current fleet of more than 300 robots around the world, we collect lessons every day to ensure a greater reliability for our upcoming generations of robots.

Why did you choose to partner with Swift Navigation?

We share a vision with Swift: “Accessible automated solutions serving sustainable goals.” We also share other important values, such as “iterate quickly” and “focus on what matters.”

March 31, 2022
Smart speakers need localization techniques

Smart speakers such as the Amazon Alexa or Google Home haven’t mastered determining user location in the home. Solving this problem was the focus of a University of Illinois at Urbana-Champaign research team’s recently published paper.

The team — led by Coordinated Science Lab graduate student Sheng Shen — explores the development of VoLoc, a system that uses the microphone array on Alexa as well as room echoes of the human voice to infer user location inside the home.

If applied, after receiving commands such as “turn on the light” or “increase the temperature,” Alexa would know which light and room is intended. Using a technique known as reverse triangulation, Shen and advisor Romit Roy Choudhury are getting closer to voice localization.

“Applying this technique to smart speakers entails quite a few challenges,” Shen said. “First, we must separate the direct human voice and each of the room echoes from the microphone recording. Then, we must accurately compute the direction for each of these echoes. Both challenges are difficult because the microphones simply record a mixture of all the sounds altogether.”

VoLoc addresses these obstacles through an align-and-cancel algorithm that iteratively isolates the direction of each arriving voice signal, and from them, reverse triangulates the user’s location. Some aspects of the room’s geometry is spontaneously learned, which helps with the triangulation.

While this is an important breakthrough, Shen and Roy Choudhury plan to expand the research to more applications. “Our immediate next step is to build to the smart speaker’s frame of reference,” Shen said. “This could mean superimposing the locations, as provided by VoLoc, on a floorplan to determine that the user is in the laundry room. Alternatively, if the smart speaker picks up the sounds made by the washer and dryer in the same location as the voice command, it can come to the same conclusion.”

Citation. S. Shen, R. Choudhury et al., “Voice Localization Using Nearby Wall Reflections,” to be presented at MobiCom 2020: 26th Annual International Conference on Mobile Computing and Networking, Sept. 21–25, London, UK.

Photo: Michael Wapp/iStock Editorial / Getty Images Plus/Getty Images

March 26, 2020
Research Roundup: Spoofing-resistant UAVs

By Alexander Rügamer, Daniel Rubino, Xabier Zubizarreta, Wolfgang Felber, Fraunhofer IIS, and Jan Wendel and Daniel Pfaffelhuber, Airbus Defense and Space GmbH

This work presents a new secure localization method that can be used for UAVs to obtain a new level of protection against hostile and unauthorized UAVs. While non-spreading code-encrypted (SCE) GNSS devices can be blocked, authorized UAVs using this method have unrestricted access to the non-spoofable and trusted SCE GNSS. The proposed method is to store short sequences of SCE PRN code chips on the user receiver before the mission.

The Precalculate & Process architecture. (Images: Fraunhofer IIS)

These SCE PRN code chips allow the user receiver to calculate at pre-defined points in time a secure and trustable SCE PVT position. Since no communication channel is required, this method mitigates the risk that hostile forces may try to jam the UAV’s radio control. Moreover, radio silence can be realized, which is beneficial or even required for some missions.

No dedicated security module required on the user terminal, no SWaP problems, no keying issues, no handling of controlled items on user side, no need for a communication link giving rise to the availability and radio silence issues, and no security issues due to the short SCE PRN code chip sequences valid only for the limited mission duration and inside a limited area.

Potential target markets for this method are police and special forces and other authorized users which are allowed to use certain SCE GNSS and would like to equip their UAVs with a secure, unspoofable positioning solution. Check out more information here.

January 18, 2019
The role of GNSS in driverless cars
Authenticated localization in driverless cars

Growing awareness of the vulnerabilities of GNSS signals — weak, unencrypted and easily jammed or spoofed — have made GNSS less important to steering the driverless vehicle. What’s up with that?

Extensive visual map databases are being created that, when coupled with cameras, radars and lidars on the vehicle and processed by artificial intelligence (AI) algorithms, enable the driverless car to be steered much the way humans drive. Pattern recognition processing in the vehicle allows it to “read” street signs and recognize landmarks, registering its position on the map.

This is the way a person drives in his or her home town, where they always know their orientation and don’t need GNSS. The AI processing “brain,” with access to huge map databases, either through local storage or a network connection, will always be in its familiar home environment: continuously knowing its own position and properly oriented for navigation.

So, will GNSS become unnecessary in the car of the future? Probably not.

First, no one method of navigation is foolproof, and today, GNSS is our primary method of navigating our cars. It is a cost-effective, accurate way of determining position in real time, and with the integration of inertial navigation sensors to handle cases when GNSS is intermittently unavailable, it is improving.

Second, it is not just the car itself that needs to know its location for navigation, but also others outside the car. Ride-sharing apps like Uber and Lyft, car-sharing, usage-based insurance apps, dynamic toll charging, and parking apps all depend on knowing where the car is at all times. GNSS offers sufficient accuracy for all these apps by providing location coordinates. Therefore, a GNSS receiver will most likely remain in the car.

The case for jamming and spoofing

Recall, however, that one of the weaknesses of GNSS is its open, unencrypted format. It is becoming increasingly easier to spoof these signals. Car-sharing, usage-based insurance and dynamic toll charging apps all create a monetary incentive for fraud that can be implemented with a spoofer. For example, a car in a car-sharing network can report a fake position indicating that it is safely parked in a secure area — while in reality, a thief is busy driving it away.

(Image: Orolia)

Let’s assume that all wireless connections to and from the car are secure. This is a reasonable assumption, although recently there have been demonstrations of carjacking via unsecure remote links. Standard SSL encryption, similar to what is used to enter credit card information on the internet, works well here. We have both the awareness and the technology now to prevent such carjackings from ever reoccurring.

However, even if communication links are secure, a GNSS spoofer in the car can fool the GNSS receiver into reporting a fake “safe” position right as it is being stolen. The same is true for insurance or toll apps. And the fraud does not have to be sophisticated. A simple, low-cost jammer can deny proper position just long enough to skirt payment. A secure location method is needed.

Other signals for localization

What would an ideal signal for localizing a driverless car look like?
- It needs to be much stronger than GNSS so it is not easily jammed.
- It needs to be encrypted so it cannot be spoofed.
- It must be ubiquitous, available worldwide.
- It must be reliable and robust — with 99.999% availability or better.
- It must be practical and priced for the mass-market automotive application.
Though accuracy is always important, the signal used for localization does not have to be as accurate as GNSS is today. Accuracy to 10s of meters is sufficient for all these applications needing fraud protection since it would not be used for steering the car, but rather, only localization. It can also be used in tandem with GNSS to authenticate a reported position when a GNSS signal is available.

Such a signal is available today, worldwide: STL (Satellite Time and Location). Carried on the Iridium satellites, it is a special purpose signal that is more than 30 dB stronger than GNSS and encrypted for anti-spoof protection. Decoding of this signal is available via a subscription model to users.

Here’s how it would work using a car-sharing example. A group of people subscribe to a car-sharing service that provides X number of cars to serve Y number of people, where X is less than Y. The service optimally schedules people when and where a car will be available. The service provider needs to know the whereabouts of the cars at all times to maximize utilization of the fleet, so every car has a GNSS receiver in it.

But to ensure the authenticity of these reports, they also have a secure localization receiver. This receiver is assigned a unique ID that is authorized to decode the encrypted signal. (Eventually, we expect this receiver and GNSS to converge into one device much the way multi-GNSS receivers operate today).

If a position report does not agree with the authentic localization report, the fleet manager can act to recover the car immediately. Insurance providers who cover secure localization-equipped cars would also give preferential rates as an anti-theft device.

(Image: Pavel Vinnik/Shutterstock.com)

Could PRS do it?

The new Public Regulated Service (PRS) from Galileo is encrypted and could provide a similar level of authentication protection, if made available. However, it is still a weak GNSS signal that can easily be jammed. Of course, any signal can be jammed, even one that is a thousand times stronger than GNSS.

However, given the robust nature of a very strong signal, the managing system that is monitoring the cars — the insurance, toll or car-sharing system, for example — can alarm upon the loss of positioning information. Such alarms on a GNSS-only car would be frequent and often erroneous due to simple fades, yielding so many false alarms that it would render the monitoring system useless. But a loss of both the strong localization signal and GNSS would likely be considered suspicious and result in a valid alarm.

GNSS navigation is truly one of the great advances of the modern era, giving us precise time and location for any place in the world. Its two major weaknesses — that it is easy to jam and spoof — can be overcome by augmenting it with other stronger encrypted signals, such as STL, providing robust jam-resistance and positive authentication.
August 1, 2018