AUTONEWS

How cellphone data can improve the travel experience at airports

Every time a traveler’s phone pings a cell tower, it leaves a tiny digital footprint.  

A new international study co-authored by a Texas A&M College of Agriculture and Life Sciences researcher shows those footprints, multiplied by millions, can help airports predict passenger surges before they happen, and cut the congestion and emissions that come with them.

The study, published in Business Strategy and the Environment, used more than 9 million anonymized mobile network signals collected around Lisbon Airport in Portugal to build a forecasting model that outperformed standard prediction methods by double-digit margins, cutting forecast errors by up to 24%.

Babak Taheri, Ph.D., professor and associate department head of graduate programs in the Arch H. Aplin III ’80 Department of Hospitality, Hotel Management and Tourism, was part of the research team, along with faculty from Molde University College in Norway and Inov Inesc in Portugal.

“Lisbon was a strong test case because it brings together many of the challenges airport managers deal with every day,” Taheri said. “It has high passenger volumes, strong seasonal peaks, a mix of domestic and international travelers, and real pressure on capacity, congestion and ground transportation.”

With global air travel expected to keep growing, researchers say anonymized mobile data could become a valuable tool for sustainable airport planning. (Sam Craft/Texas A&M AgriLife)

From data to decisions...The team tracked anonymized, aggregated signals from mobile devices across 119 grid cells covering the airport’s footprint over a full year, then fed the patterns into a forecasting model built on Prophet, a time-series tool suited to capturing seasonal swings and holiday effects. Any grid-and-time slice with fewer than 10 devices was excluded, and no individual users could be identified.

What sets the study apart is what happens after the forecast. The research team translated those projections directly into potential operational decisions. Their forecast showed how the Lisbon Airport could use the information to determine how many security lanes to open, how many staff to schedule and how much to shorten the intervals between metro trains.

Babak Taheri, Ph.D., professor and associate department head of graduate programs in the Arch H. Aplin III ’80 Department of Hospitality, Hotel Management and Tourism, with a student in department’s Digital Transformation Lab, is researching how anonymous cellphone data can help airports anticipate passenger surges, improve operations and reduce emissions. (Michael Miller/Texas A&M AgriLife)

Under the scenarios modeled, better-coordinated staffing and transit schedules could reduce ground-transport congestion and associated emissions by 14-26% during peak travel periods.

“Telling an airport manager that passenger demand may increase by a certain percentage is only partly useful,” Taheri said. “Telling them what that could mean for staffing or the number of security lanes makes the information much more actionable.”

With global air travel expected to keep growing, researchers say anonymized mobile data could become a valuable tool for sustainable airport planning--image above (Sam Craft/Texas A&M AgriLife)

The model isn’t perfect. For example, it underestimated actual passenger counts during a June 2023 spike, a gap that Taheri said underscores why forecasts should inform planning ranges rather than serve as a single guaranteed number.

“There will always be days or periods when actual demand moves outside expectations,” he said.

International passenger flows also proved harder to predict than domestic ones, likely because they’re shaped by holiday calendars, economic conditions and other factors across multiple countries.

This finding points to combining mobile data with flight schedules and weather data for sharper forecasts, Taheri said.

A model for other hubs...With global air travel projected to grow by the billions over the next two decades, the researchers argue mobile network data is an underused asset for sustainability-minded airport management.

Taheri cautioned, however, that the approach isn’t plug-and-play.

“Every airport has its own passenger mix, terminal layout, transfer traffic and seasonal patterns, so the model would need to be trained and calibrated using local data,” he said.

The staffing and security thresholds outlined in the study haven’t yet been tested in live operations.

“The current study has not yet been deployed as a live airport operating system,” Taheri said. “What we have demonstrated is that the data and forecasting can be translated into decisions airport managers recognize and act on. The logical next step is an industry pilot where we test those decisions in real time, measure what works and refine the system with airport operators.”

If validated in real-world operations, the approach could give airports a new tool to anticipate pressure points before terminals and transportation network become congested.

Cellphones improve the airport travel experience by providing anonymized location footprints and real-time connectivity that allow operators to predict crowd surges, optimize staffing, and eliminate terminal friction.According to a September 2026 international study co-authored by researchers at Texas A&M University, cell tower pings from passenger phones can reduce forecasting errors by up to 24%. This helps airports handle congestion before it bottlenecks the traveler's journey

Predictive crowd management & logistics:

-Surge forecasting: By analyzing millions of anonymized network signals, airports can anticipate passenger volume hours in advance.

-Actionable staffing: Instead of guessing traffic, managers can use data to calculate exactly how many security lanes to open.

-Transit coordination: Better-aligned bus, taxi, and train schedules can cut ground-transport congestion by 14% to 26% during peak hours

Streamlining the terminal experience:

-Digital identities: Travelers can bypass traditional checkpoints using securely stored mobile biometrics and state-issued digital IDs via platforms like Apple Wallet or Google Wallet, authorized by the TSA

-Virtual queuing: Some hubs now use cellphone location data and apps to let passengers reserve a spot in line virtually, freeing them from standing in physical queues.

-Indoor navigation: Bluetooth beacons push personalized, real-time wayfinding maps and localized store offers directly to a passenger's smartphone based on their exact terminal location

Behind-the-scenes airport operations:

-Faster aircraft turnaround: Cellular-connected systems immediately beam diagnostic data from landing aircraft to ground crews, speeding up maintenance and reducing flight delays

-High-density 5G network support: Next-generation 5G connectivity keeps passenger transactions, boarding pass scans, and real-time gate updates moving seamlessly even inside heavily crowded terminals.

Texas A&M College of Agriculture and Life Sciences researcher

AUTONEWS

A third Dacia model for the C-segment is off the table

The Dacia Bigster hit the market last year and proved to be a hit in the C-SUV segment. In fact, the Renault Group brand expects the C-segment to account for one-third of its sales by 2030, despite having only recently entered the market.

This goal was reiterated last March in a statement describing the Bigster and the upcoming Striker as a "perfectly complementary duo," featuring distinct personalities yet sharing the same DNA.

The firm began its push into Europe's biggest market (often referred to as the 'mid-sized' segment) in 2024 with the launch of the Nissan Qashqai-sized Bigster SUV, and is soon to expand its presence in the category with the closely related, estate-shaped Striker.

This move into larger cars has been a highly successful one: the Bigster ranked as Europe's best-selling C-SUV in the second half of last year and Dacia anticipates that the segment will account for a third of its sales by 2030.

However, with the firm now committed to launching four new EVs by 2030 (beginning imminently with the new Spring city car), it has scrapped plans to add a third petrol-powered C-segment model.

The Striker was originally planned to be the second of these 4.5m-long models, with the closely related third entry due to follow next year. 

But now Dacia CEO Katrin Adt has poured cold water on the prospect of new mid-sized petrol cars in the immediate future, citing EVs and smaller cars as the priority.

"We have a very successful entry with the Bigster, we see the potential for the Striker and currently we don't have any other plans,” she told Autocar when asked for an update on the next car.

Unveiled nearly six months ago, the Striker is set to launch soon—larger than the Bigster and featuring a wagon-like body style. The brand's commitment to C-segment models stops there; there are no plans to expand the lineup with a third model, an idea that had previously been under consideration.

Executive Katrin Adt told *Autocar* that she is confident in the potential of the Bigster and Striker, noting there are no plans for another model in that segment. According to Adt, the focus is on "the next-generation Sandero," while also pointing out that Dacia is not known for having an extensive model range.

Originally introduced in 2008, the Sandero is heading into its fourth generation. It is a subcompact hatchback that is expected to offer multi-energy powertrain options in the future, including fully electric versions. Meanwhile, a new Spring—developed in tandem with the Renault Twingo E-Tech—is in the works, and the Hipster, a microcar concept, was unveiled last year.

MOTO GP

Marc Márquez's masterclass at Misano

Spaniard Marc Márquez claimed victory in the San Marino Grand Prix Sprint race, demonstrating great speed and track control as he finished ahead of Italian Marco Bezzecchi. In doing so, both riders narrowed the gap to the 2026 MotoGP leader, Spaniard Jorge Martín, who finished third.

With Marc Márquez and Jorge Martín opting for the medium rear tire—unlike the soft tires chosen by the rest of the front row—the lights went out and Márquez managed to overtake Bezzecchi. Fabio Di Giannantonio slotted in behind Bezzecchi after passing Fermín Aldeguer a few corners later; Martín took advantage of this to overtake Aldeguer, who also lost a position to his compatriot Pedro Acosta.

The race settled into a rhythm with no changes in position until the third lap, when Aldeguer made contact with Acosta; Álex Márquez capitalized on the situation to overtake both and move into fifth place. This created a gap between the top four and the rest of the pack. Raúl Fernández overtook Aldeguer, who then had to defend eighth place against Luca Marini. Just behind them, Fabio Quartararo crashed while attempting to overtake Diogo Moreira, who miraculously managed to stay upright.

Marc Márquez dropped his lap time below 1:30 to set the fastest lap and build a seven-tenth lead over Bezzecchi. Bezzecchi, in turn, was pulling away from 'Diggia' and Martín, who were being dangerously closed in on by Álex Márquez—who had also managed to dip below the 1:31 mark.

Marc Márquez stabilized his lap times and extended his lead to one second by the sixth lap, while behind him, Martín snatched third place from Di Giannantonio. At the front, the trend shifted on the eighth lap as Bezzecchi cut Marc's lead by four-tenths of a second, bringing the gap down to seven-tenths.

On the following lap, he shaved another tenth off the deficit, with four laps still remaining. On the following lap, the gap shrank to just half a tenth, setting up a final three-lap dash with little more than half a second separating them. But then Marc Márquez pushed hard to set the fastest lap and open up a further one-second lead, securing the victory.

Marco Bezzecchi had to settle for second place, and Jorge Martín rounded out the podium after pulling away from Fabio Di Giannantonio and Álex Márquez; Pedro Acosta finished sixth ahead of Raúl Fernández and Fermín Aldeguer, while Luca Marini took the final point on offer.

Bezzecchi started from the front, but Márquez fought back... Despite using the soft rear tire—while Bezzecchi opted for the medium compound—Márquez got a better start. The Ducati rider took the lead entering the first corner and tried to pull away right from the opening meters.

However, Bezzecchi didn't let his rival get away. The Aprilia rider began pressuring Márquez and narrowed the gap during the race. Even so, the Spaniard responded during the decisive moments and managed to keep his pursuer at bay to secure the victory.

Márquez took the checkered flag after 13 laps, finishing 0.743 seconds ahead of Bezzecchi. Jorge Martín, the championship leader and also an Aprilia rider, rounded out the podium, 1.794 seconds behind the winner.

More details...The race began with Marc Márquez taking the lead at the first corner, leaving Marco Bezzecchi in second place, followed by Fabio Di Giannantonio and Jorge Martín. Fermín Aldeguer dropped to fifth place. A little further back, Diogo Moreira moved up to tenth place during the opening corners of the race.

Still on the first lap, Pedro Acosta moved up to fifth, overtaking Aldeguer, who then came under pressure from his Gresini teammate, Álex Márquez. In the pack behind, Francesco Bagnaia passed Pol Espargaró, taking 14th place. Álex Márquez overtook both Aldeguer and Acosta, breaking into the top five by the third lap.

Aldeguer plummeted down the order to eighth place, while Fabio Quartararo crashed his Yamaha following an incident with Moreira, who continued in the race. The Brazilian lost a position to Johann Zarco, dropping to 11th place. At the front, Di Giannantonio received a track-limits warning, while Márquez opened up a one-second gap over Bezzecchi.

On the seventh lap, Martín attacked Di Giannantonio and successfully overtook him, taking third place. In the battle for the lead, Bezzecchi once again closed the gap to Márquez to seven-tenths of a second by the ninth lap. Meanwhile, Moreira held onto 11th place.

With four laps remaining, Bezzecchi cut the gap to Márquez to half a second, while Martín lacked the pace to close in on the top two. Di Giannantonio, meanwhile, came under pressure from Álex Márquez in the fight for fourth place.

The race entered the final lap with Márquez 1.1 seconds ahead of Bezzecchi, cruising toward victory. Bezzecchi and Martín rounded out the top three, while Moreira finished 11th.

Fabio Di Giannantonio finished in P4, ahead of Álex Márquez and Pedro Acosta. Raúl Fernández took P7, while Fermín Aldeguer finished in P8.

Luca Marini secured P9, and Johann Zarco rounded out the top 10. Diogo Moreira finished in P11, 10.493 seconds behind Márquez.

Franco Morbidelli finished in P12, followed by Francesco Bagnaia. The Italian Ducati rider had a difficult race and crossed the finish line 17.763 seconds behind the winner.

Fabio Quartararo had a tough race; the Yamaha rider crashed early in the Sprint while attempting to overtake Diogo Moreira for P11.

Quartararo managed to rejoin the race but lost a lot of time due to the incident. Consequently, he finished in P21—last place—24.419 seconds behind Márquez.

Brad Binder finished P14, followed by Enea Bastianini and Pol Espargaró. Toprak Razgatlioglu finished in P17, while Jack Miller took P18. Álex Rins and Joan Mir rounded out the classification in P19 and P20

by: Autonews

TESLA

Tesla pairs Cybertruck PowerShare with Powerwall 3 for extra days of home backup in an emergency

Tesla has enabled Cybertruck PowerShare integration with its Powerwall panels, allowing the truck’s 123 kWh battery to extend emergency home backup. The feature adds runtime rather than output, retaining an 11.5 kW system cap.

Tesla has finally flipped the switch on Cybertruck PowerShare working together with Powerwall as a home backup in an emergency.

The feature was promised back in November 2023 alongside the Cybertruck itself, and it only went live this week, exclusively for Powerwall 3 owners who fill an opt-in survey. It is coming to Powerwall 2 and Powerwall+ owners later this year, too.

The pitch from Tesla is that the Cybertruck can stretch a home's backup time by more than three days, which Tesla frames as adding "9 additional Powerwalls" worth of capacity. That math checks out, at least on paper, since Tesla typically assumes 30 kWh of daily household use, and the Cybertruck's 123 kWh battery capacity is indeed close to nine Powerwall 3 units stacked together.

What actually changes versus the setup that Cybertruck owners already had is that now the electric truck can be treated as the sole backup source. Since 2024, PowerShare could power a home directly through a separate PowerShare Gateway and Tesla's Universal Wall Connector that can be had on Amazon, with no Powerwall required.

The new pairing is different and stricter, as it requires a Powerwall 3, a Gateway 3, or a Backup Gateway 2 or Backup Switch, and it explicitly forbids a PowerShare Gateway on the same system. In this mode, the Powerwall discharges first, and the truck kicks in as reserve capacity once needed, even topping the Powerwall back up while the grid is down.

Still, total output stays capped at 11.5 kW combined, not per device, and the Cybertruck only adds runtime, not extra simultaneous power. In other words, running the AC and the EV charger at once during an outage would not be possible, but hooking up the Cybertuck will add days of continuous power to one's home.

What is powershare...Powershare bidirectional charging technology turns equipped Tesla vehicles into mobile power banks, enabling energy to flow to and from the battery. For Cybertruck, this means a fully charged vehicle can power an average home for over three days1 during an outage with Powershare Home Backup. With Powershare Grid Support, stored energy can also be sent back to the grid, earning owners compensation in return.

For Powershare to work, the energy stored in the battery of Cybertruck needs to be converted from direct current (DC) to alternating current (AC) power, which is how both home appliances and the grid are powered. This conversion from DC to AC is done through the vehicle's built-in inverter. That means owners only need to add Powershare Gateway and Universal Wall Connector to use Powershare, keeping installation simple and costs low.

Powershare Home Backup, Tesla’s vehicle-to-home (V2H) technology, provides backup power to homes using the energy stored in the battery of Cybertruck.

When a grid outage occurs and your vehicle is plugged in, Powershare Gateway automatically detects loss of power and signals Cybertruck to begin discharging, allowing the home to switch from grid power to vehicle power within one minute2. With Universal Wall Connector, Cybertruck can deliver up to 11.5 kW of continuous power to keep lights, appliances, Wi-Fi and other critical systems running. It is a cleaner and quieter alternative to traditional generators, requiring no fuel storage, regular maintenance or manual operation.

To extend backup duration, owners can pair Cybertruck with Tesla Solar Panels to recharge their vehicle’s battery during the day.

Powerwall...Powershare Home Backup will soon support Powerwall. Once available, your plugged-in Cybertruck will be able to add extra energy storage to extend Powerwall backup time by three days while Powerwall helps run high-demand appliances, no additional equipment required. When your Cybertruck is on the road, Powerwall continues to provide your home with backup electricity.

In addition to powering homes, Powershare Grid Support, Tesla’s vehicle-to-grid (V2G) technology, enables Cybertruck to send stored energy to the grid during periods of high demand, earning owners compensation for contributing to their community's energy resilience. Cybertruck can determine when to recharge and send energy back to the grid based on market conditions and the owner's system preferences.

Powershare Grid Support is currently available to owners in parts of California and Texas3 and requires no additional equipment beyond what's already installed for Powershare Home Backup. Owners who wish to enable Powershare Grid Support should inform their Tesla Certified Installer to ensure the installation meets local electrical codes for safely exporting power back to the grid.

In California, Powershare Grid Support allows owners to participate in virtual power plants4 (VPPs), which are utility-powered programs designed to support grid stability when electricity demand fluctuates. These programs are available through the Emergency Load Reduction Program5 (ELRP) for customers of Pacific Gas and Electric Company (PG&E), Southern California Edison (SCE) and San Diego Gas and Electric (SDG&E).

Between May and October, owners can expect to support 30 to 60 hours of grid events per year, with each event lasting one to three hours.6 

Participation in VPPs can earn owners up to $690 in cash per year, calculated at $2 for every kWh their vehicle delivers back to the grid during an event.

Tesla App...Both Powershare Home Backup and Powershare Grid Support can be enrolled in, tracked and managed through the Tesla app, providing real-time visibility into energy usage and backup power status.

For owners who want to test their setup by simulating a power outage, Go Off-Grid mode temporarily disconnects the home from the grid, even if power is available. When activated, the home runs only on the battery power provided by Cybertruck and solar energy (if available). The Tesla app monitors remaining backup duration and automatically reconnects to the grid if energy demand exceeds system capacity.

Powershare is designed to work around the owner's driving needs. Through the Tesla app, owners can set a minimum charge level, ensuring the Cybertruck always retains enough range for daily use. Neither Powershare Home Backup nor Powershare Grid Support will require the vehicle to discharge below this threshold.

Owners can also schedule their departure time to ensure their vehicle is fully charged and ready when needed. When grid events occur, owners are notified and may choose to participate or opt out.

Participating in Powershare Grid Support does not impact vehicle warranties. Tesla's battery management system ensures that all charging and discharging occur within optimal parameters.

Autonews

FORD

The Ford Ranger Super Duty 6x6 is a contender for the British Army's new vehicle fleet

Just a few months ago, Ford revealed that military and government agencies across Europe and North America were considering adapting commercial pickup trucks for defense operations. One such example is the mid-size Ford Ranger, which has been proposed for the British Ministry of Defence's Light Multi-Role Vehicle (LMV) program.

A militarized version of the Ford Ranger—featuring six wheels and a hybrid powertrain—was previously developed through a collaboration between the engineering firm Ricardo and Ford. Now, a new company has joined the effort: the British subsidiary of General Dynamics Land Systems. This division of the U.S. aerospace and defense giant General Dynamics Corporation specializes in land combat and tactical vehicles.

The company pledged that the three partners would deliver a "proven, scalable, and cost-effective solution," adding that their modified vehicle "meets the British Army's requirements while creating significant export opportunities and maximizing value for British taxpayers."

The British Ministry of Defence aims to acquire approximately 2,500 Light Multi-Role Vehicles to replace its aging fleet of Land Rover Defenders and Steyr-Puch Pinzgauers by 2030, as part of a £4.8 billion initiative. According to *Defense Blog*, citing the specialist publication *Army Technology*, vehicle evaluation trials are scheduled to take place between October 2026 and January 2027. The modified Ford Ranger is one of the contenders, alongside the Land Rover Defender Wolf Series II and other proposals from General Motors and Ineos.

Photos released by General Dynamics on social media, as well as by Ford on its "From The Road" blog, show two Ford Ranger units with varying levels of modification. The first pickup appears to be based on the Wildtrak trim level and features a front bumper brush guard, a sports bar behind the cab, and a custom military-style wrap.

The second version is more interesting as it features six wheels. Although the prototype's bodywork resembles that of the entry-level XLT trim, the chassis is derived from the more capable Ford Ranger Super Duty developed in Australia.

Under the hood lies a 3.0-liter V6 turbodiesel engine producing 209 hp and 600 Nm of torque, paired with a hybrid system. Ricardo previously released details regarding an electric motor mounted on the rear axle, which contributes 286 hp in both 6x6 and 6x4 configurations.

The pickup is equipped with steel underbody protection, a robust transfer case, front and rear differential locks, and a reinforced suspension system; this setup provides nearly 300 mm of ground clearance and allows it to ford water up to 850 mm deep.

It also boasts a towing capacity of 4,500 kg, a payload capacity of 1,982 kg, and a gross combination weight rating (GCWR) of 8,000 kg. Finally, the interior features third-party tactical mission software integrated into the standard 12-inch infotainment display. The automaker confirmed that the turbodiesel engine is manufactured in London and the hybrid system in Halewood. Ford's facilities in Dunton house the commercial vehicle development, technical development, and conversion engineering departments, while Daventry serves as the central logistics hub for parts logistics and distribution.

 

by Autonews

DOSSIER

AUTONEWS

U4D: Uncertainty-aware 4D world modeling from LiDAR sequences

Modeling dynamic 3D environments from LiDAR sequences is central to building reliable 4D worlds for autonomous driving and embodied AI. Existing generative frameworks, however, often treat all spatial regions uniformly, overlooking the varying uncertainty across real-world scenes. This uniform generation leads to artifacts in complex or ambiguous regions, limiting realism and temporal stability. In this work, we present U4D, an uncertainty-aware framework for 4D LiDAR world modeling. Our approach first estimates spatial uncertainty maps from a pretrained segmentation model to localize semantically challenging regions.

Modeling dynamic 3D environments from LiDAR sequences is fundamental to constructing reliable 4D world models that enable autonomous systems to perceive, simulate, and interact with the physical world over time. LiDAR provides precise geometric and depth information, forming the basis for perception, mapping, and planning in autonomous driving, robotics, and 3D scene reconstruction. However, collecting large-scale, diverse, and annotated LiDAR data remains costly and labor-intensive, motivating increasing interest in generative LiDAR modeling for scalable simulation, data augmentation, and pretraining.

Recent advances have explored LiDAR scene generation via adversarial, variational, and diffusion-based generative frameworks . Early efforts focus on object-level point clouds, whereas recent methods such as LiDARGen, LiDM, and LiDARCrafter synthesize large-scale and even dynamic LiDAR scenes. Despite these advances, existing methods treat spatial regions equally during generation, ignoring the varying semantic difficulty of real-world data. This uniform assumption often causes unreliable reconstruction in geometrically or semantically complex regions, such as thin poles, moving objects, and distant surfaces, where predictive confidence is low.

We observe that reliable LiDAR world modeling requires understanding the underlying uncertainty itself. Real LiDAR observations exhibit inherently non-uniform difficulty: while some areas are well-defined, others – such as occluded areas, small-scale structures, or semantically ambiguous regions – remain uncertain. Ignoring this asymmetry leads to geometric artifacts and temporal instability. Inspired by how humans resolve ambiguous regions before perceiving global context, we propose to model uncertainty explicitly, generating difficult regions first as structural anchors for the rest of the scene.

To this end, we propose U4D, an uncertainty-aware framework for 4D LiDAR world modeling. U4D leverages spatial uncertainty as a structural prior to guide scene generation. Our framework first estimates an uncertainty map from a pretrained LiDAR segmentation network to localize semantically ambiguous or geometrically unreliable regions. It then performs two sequential stages of generation: An uncertainty-region diffusion stage, which focuses on reconstructing high-entropy regions with fine geometric fidelity, and an uncertainty-conditioned completion stage, which synthesizes the remaining areas conditioned on these reconstructed structures. The two stages share a unified latent scene representation, enabling global contextual cues to refine local uncertainty and ensuring geometric and temporal consistency across the generated 4D scenes.

To further ensure stable temporal evolution, U4D integrates a Mixture of Spatio-Temporal (MoST) block, which explicitly decomposes and adaptively fuses spatial and temporal representations within the diffusion process. This design enables the generation of LiDAR sequences that are both geometrically precise and temporally coherent. Extensive experiments on the nuScenes and SemanticKITTI datasets demonstrate that U4D consistently outperforms existing LiDAR generation frameworks in terms of geometric fidelity, temporal stability, and downstream generalization

We propose U4D, the first uncertainty-aware generative framework for LiDAR scene synthesis. U4D generates LiDAR scenes in a “hard-to-easy” manner through two sequential stages. It estimates an uncertainty map from a real scan using a pretrained segmentation model and employs an unconditional diffusion process to reconstruct high-fidelity uncertain regions. Conditioned on these reconstructed areas, U4D completes the remaining scene to ensure structural integrity and global coherence. Both stages share latent representations, enabling global context to refine local uncertainty. To maintain temporal stability, U4D integrates a Mixture of Spatio-Temporal (MoST) block within the diffusion backbone. The MoST block decouples and adaptively fuses spatial and temporal features, enabling the generation of LiDAR sequences that are both geometrically precise and temporally coherent.

Uncertainty measurement in 3D...Real-world LiDAR scenes exhibit non-uniform difficulty across spatial regions. Some structures, such as ground or buildings, are geometrically stable and semantically consistent, while others are inherently uncertain due to factors such as distance-induced sparsity, occlusion, small-scale objects, or semantic ambiguity between visually similar categories. These uncertainty-prone regions frequently appear at long ranges, around object boundaries, or in areas of low point density, leading to inconsistent predictions in safety-critical perception tasks. Explicitly identifying and modeling these regions allows generative models to focus first on structurally unstable and perceptually ambiguous areas before extending to the entire scene, thereby producing more realistic priors and improving downstream robustness.

Uncertainty measurement in 3D...Real-world LiDAR scenes exhibit non-uniform difficulty across spatial regions. Some structures, such as ground or buildings, are geometrically stable and semantically consistent, while others are inherently uncertain due to factors such as distance-induced sparsity, occlusion, small-scale objects, or semantic ambiguity between visually similar categories. These uncertainty-prone regions frequently appear at long ranges, around object boundaries, or in areas of low point density, leading to inconsistent predictions in safety-critical perception tasks. Explicitly identifying and modeling these regions allows generative models to focus first on structurally unstable and perceptually ambiguous areas before extending to the entire scene, thereby producing more realistic priors and improving downstream robustness.

Uncertainty-conditioned scene completion...While sparse uncertainty scenes emphasize semantically ambiguous regions or structurally unstable regions, they only capture partial geometric information. To generate complete and coherent LiDAR frames, we design an uncertainty-conditioned diffusion model that synthesizes full scenes under the guidance of these uncertainty priors.

Mixture of spatio-temporal...Temporal coherence is essential for modeling dynamic real-world environments. While prior methods  focus on spatial reconstruction, maintaining consistent temporal evolution across frames remains challenging in dynamic modeling. To jointly ensure spatial fidelity and temporal consistency, we propose the Mixture of Spatio-Temporal (MoST) block. MoST is a core diffusion component that simultaneously captures fine-grained spatial geometry within each frame and smooth temporal transitions between frames, producing LiDAR sequences that are both geometrically accurate and temporally stable.

Comparative study...Scene-Level Fidelity. We first benchmark U4D against state-of-the-art LiDAR scene generation methods from a scene-level spatial fidelity perspective. Specifically, we sample 

 sequences and evaluate the first frame of each sequence to ensure a fair comparison with single-frame generation baselines producing the same number of scenes. As shown in Tab. 1 and Tab. 2, U4D consistently outperforms existing methods, achieving an FRD of 

For BEV-based metrics, including JSD and MMD, U4D also achieves competitive or superior performance, demonstrating robust spatial consistency across viewpoints. These results highlight U4D’s strong ability to generate geometrically accurate and perceptually consistent LiDAR scenes.

Table below:Comparison of state-of-the-art LiDAR scene generation methods on the SemanticKITTI [5] dataset. Metrics marked with  ↓ indicate that lower values are better. The MMD scores are reported in units of 10−4 The best and second-best scores are highlighted in bold and underline, respectively.

Method	Venue	FRD $\downarrow$	FPD $\downarrow$	JSD $\downarrow$	MMD $\downarrow$
LiDARGen [124]	ECCV’22	$735.49$	$119.69$	$0.13$	$21.90$
LiDM [83]	CVPR’24	-	$496.78$	$0.08$	$9.20$
R2DM [73]	ICRA’24	$262.85$	$12.06$	$\mathbf{0.03}$	$0.89$
Text2LiDAR [99]	ECCV’24	$567.47$	$16.78$	$0.08$	$4.24$
U4D	Ours	$\mathbf{245.73}$	$\mathbf{10.92}$	$0.04$	$\mathbf{0.85}$

Maintaining temporal coherence is crucial for sequential LiDAR generation, as inconsistent frame-to-frame predictions can lead to unrealistic scene dynamics. We evaluate U4D against recent methods, including UniScene, OpenDWM, and LiDARCrafter, on sequences sampled at 2Hz. TTCE measures deviations between predicted and ground-truth transformations via point cloud registration, while CTC computes Chamfer distances between consecutive frames. U4D consistently achieves the lowest TTCE scores across all frame intervals and maintains competitive CTC scores, reflecting its ability to generate temporally stable sequences with smooth and realistic motion patterns. These results highlight the effectiveness of the MoST block in capturing both spatial and temporal dependencies within sequences.

Design of MoST Block...The Mixture of Spatio-Temporal (MoST) block serves as a key component of our diffusion backbone, designed to adaptively fuse spatial and temporal cues for coherent LiDAR scene generation. In this ablation, we investigate how different fusion strategies affect generation quality. We first follow prior video generation approaches  and apply spatial and temporal operations in a cascaded manner. This configuration yields suboptimal results, likely because the deeper cascaded structure hampers convergence and optimization. We then decompose features into spatial and temporal branches and fuse them in parallel using either element-wise addition or concatenation. Both strategies significantly improve generation quality, as they expand network width rather than depth, facilitating more stable optimization. Finally, we introduce an adaptive fusion mechanism inspired by the mixture-of-experts paradigm, where the model learns to dynamically balance spatial and temporal information. This design achieves the best generation quality, demonstrating the effectiveness of adaptive spatio-temporal fusion for coherent 4D LiDAR generation.

To further examine how the MoST block fuses spatial and temporal information across different network stages, we analyze the relative activation weights of its two branches. As shown in the top-right of we visualize the averaged weighting distribution of spatial and temporal branches throughout the diffusion network. We observe that near the input and output layers, the spatial branch contributes more prominently, as these stages mainly focus on reconstructing local geometric details and structural integrity of LiDAR frames. In contrast, the temporal branch exhibits stronger activations in intermediate layers, where the model captures motion dynamics and ensures temporal consistency across frames. This observation validates our design intuition that spatial cues dominate at the boundaries for geometric fidelity, while temporal cues become essential in the latent space to model scene evolution and motion continuity. The adaptive allocation of activations allows MoST to balance geometric reconstruction and motion modeling, enabling the diffusion network to generate LiDAR sequences that are both spatially accurate and temporally coherent.

 

by Autonews

AUTONEWS

Mercedes-AMG CLE 646 Spezialanfertigung

The Mercedes-AMG CLE 646 Spezialanfertigung was developed with a single, clear focus: driving dynamics. This high-performance, road-legal coupé delivers an uncompromising driving experience on winding roads and closed race tracks.

To achieve this, the vehicle combines a 646 hp (475 kW) V8 biturbo engine with exclusive rear-wheel drive, significantly reduced weight, and a chassis configuration engineered specifically for precision.

A unique wide-body concept—featuring a wider track, numerous carbon-fiber components, and purpose-built aerodynamics—further enhances the vehicle's dynamic potential. To realize this, every unit of the CLE 646 "Spezialanfertigung" (special build) undergoes extensive modification at AMG headquarters in Affalterbach, adhering to the highest standards of hand-crafted production.

At its heart lies the significantly upgraded AMG M177 EVO 4.0-liter V8 biturbo engine. Its 646 hp (475 kW) output is reflected in the model's designation. Peak torque stands at 850 Nm, available across a broad engine speed range of 2,500 to 4,500 rpm.

One of the most significant modifications concerns the drivetrain. The AMG Performance 4MATIC+ system from the original vehicle has been completely removed, meaning the CLE 646 Spezialanfertigung transmits power exclusively to the rear axle. Switching to rear-wheel drive reduces weight while fundamentally altering the car's dynamic character. An electronically controlled rear limited-slip differential optimizes torque distribution between the rear wheels via variable locking rates, depending on requirements. This enhances traction, driving stability, and overall vehicle dynamics. Under appropriate driving conditions, the system also enables controlled drifting.

The standard AMG SPEEDSHIFT TCT 9G transmission delights performance-oriented drivers with its dynamic gear-shifting characteristics and an emotive rev-matching function. Depending on the selected drive program, it delivers either sporty, rapid gear changes or exceptionally smooth, almost imperceptible shifts. The AMG SPEEDSHIFT TCT 9G transmission also combines efficiency with dynamic performance.

The transmission's overall character can be tailored via the AMG DYNAMIC SELECT system, which offers a range of drive programs with specific powertrain settings.

The Launch Control system in the Mercedes-AMG CLE 646 Spezialanfertigung has also been specifically adapted for the rear-wheel-drive configuration. This new model from the Mythos series accelerates to 200 km/h in 11.4 seconds. Its top speed is 310 km/h.

The suspension system has also been specially optimized for the Mercedes-AMG CLE 646 Spezialanfertigung, having been jointly developed by Mercedes-AMG and KW automotive. The configuration allows for extremely precise adjustment based on driving style, track characteristics, personal preferences, and the vehicle's intended use. The high-performance AMG ceramic composite braking system is engineered for extreme loads and delivers consistently high braking performance, even during intense track driving. On the front axle, 420 x 40 mm ceramic brake discs are paired with six-piston fixed calipers. On the rear axle, the system features 360 x 32 mm ceramic brake discs.

Compared to a conventional braking system, this configuration reduces overall weight. This directly contributes to improved driving dynamics and more precise turn-in response. Gold-painted brake calipers further underscore the high-performance nature of the system.

The bodywork has been designed entirely with a focus on maximizing driving dynamics and performance. A comprehensively modified body structure increases rigidity and forms the foundation for the vehicle's exceptional performance potential. Numerous standard components have been replaced by solutions specifically developed for demanding road and track use. A key element of the concept is the vehicle's significantly wider stance; to achieve this, the bodywork has been reinforced and widened. The result is an increase in vehicle width of 60 millimeters at both the front and rear. The wider track improves stability, enhances cornering potential, and allows for the use of exceptionally wide wheels and tires. Carbon fiber has been used for the widened front fenders with integrated air vents, the redesigned front bumper featuring large cooling air intakes, the side skirts, the hood with a central cold-air intake, the roof, and a newly developed rear bumper with a diffuser. The reinforced trunk lid and the manually adjustable rear wing—mounted on two trapezoidal supports—are also made of carbon fiber. The rear side windows and the rear window are made of polycarbonate, while a lighter 12-volt starter battery further contributes to reducing the overall weight.

The interior reflects the same unwavering focus on lightweight construction and driving dynamics. The rear seats and sound-insulation materials have been removed. Comfort features have been minimized, and the audio system simplified to save weight. Omitting certain driving modes and driver-assistance systems also allowed for a reduction in the number of controls on the AMG Performance steering wheel. Featuring DINAMICA microfiber upholstery and a flat-bottomed shape inspired by motorsport, the steering wheel offers exceptional control.

Ultra-lightweight, high-performance racing seats with pronounced lateral support come as standard, alongside a roll-over bar that increases torsional rigidity and further enhances driving dynamics. Lightweight carbon-fiber door panels complete the interior. As part of the optional Performance package, six-point seatbelts firmly hold the driver and front passenger in place.

Vehicle ride height has also been optimized for performance, with the vehicle lowered by 22 millimeters at the front axle and 16 millimeters at the rear.

To strike the best balance between everyday usability and track performance, the front diffuser can be manually adjusted to one of two positions—"Street" or "Race"—using three fasteners, with the latter setting intended exclusively for track driving. When set to the "Race" position, the front diffuser creates a specific aerodynamic profile beneath the vehicle's front end. Its shape, resembling an inverted wing, accelerates airflow under the car and generates a Venturi effect, thereby increasing downforce on the front axle. The benefits are immediately apparent in the steering response; the Mercedes-AMG CLE 646 Spezialanfertigung remains exceptionally precise and stable at high speeds, while offering agile turn-in and clear steering feedback under high lateral loads. In conjunction with the extended front diffuser, the center section of the carbon-fiber rear spoiler can be manually adjusted to further increase rear-axle downforce during track driving and achieve the desired aerodynamic balance.

Designed for use on closed circuits with ESP deactivated, the AMG TRACTION CONTROL system allows drivers to individually adjust the level of support when managing high engine power, without intervention from the ESP stability system. The system offers precise adjustability, allowing for a configuration tailored to tire temperature and the driver's desired proximity to the handling limits. Specially calibrated for the Mercedes-AMG CLE 646 Spezialanfertigung, AMG TRACTION CONTROL allows the driver to choose from nine levels of permissible wheel slip on the driven rear axle. In pursuit of a more direct driving experience, the number of infotainment functions and other features has been reduced. Consequently, the CLE 646 Spezialanfertigung dispenses with numerous comfort and driver-assistance systems, including the PARKTRONIC parking aid.

Each vehicle is handcrafted in Affalterbach based on the CLE 53 platform. This continues the long AMG tradition of exclusive hand-craftsmanship, responsible for creating legendary models such as the AMG 300 CE 6.0 "The Hammer."

The Mercedes-AMG CLE 646 Spezialanfertigung is the second model in the ultra-exclusive Mercedes-Benz Mythos series, dedicated to rare collector's vehicles produced in strictly limited numbers. Following the Mercedes-AMG PureSpeed—produced in a global run of just 250 units and featuring an open-top, two-seater configuration without a roof or windshield—the Mercedes-AMG CLE 646 Spezialanfertigung now joins the series. Street-legal and produced in a limited run of only 30 units, this vehicle places an uncompromising emphasis on track performance. Combining striking looks, top-tier performance, and an exceptionally direct driving experience, the model has become a highly coveted product—so much so that it sold out even before its world premiere.