Chinese EV Tech Surges in Autonomy and Smart Roads

Chinese EV and autonomous driving news this week points to a clear industry shift: competition is no longer just about selling more electric cars, but about proving real-world driver-assistance capability, scaling robotaxi operations overseas, and building the smart-road infrastructure that supports both. In China, D1EV’s latest intelligent driving rankings show GAC Aion closing to within 0.28 points of long-time leader Exeed, while Huawei-backed Aito made a strong debut. At the same time, NavInfo secured a RMB 100.26 million smart-city roadside perception project in Beijing, and Pony.ai expanded its seventh-generation Robotaxi program to Manchester less than a week after announcing London. Together, these developments show how Chinese mobility players are competing across vehicles, software, infrastructure, and global deployment.

D1EV Smart Driving Rankings: The Margins Are Now Tiny

The most revealing takeaway from D1EV’s September intelligent driving competition rankings is how little now separates the front-runners. In the Shenyang final, 4 of the 11 finalist vehicles scored above 90 points. Over the rolling six-month ranking window, the gap between first and second place narrowed to just 0.28 points, while the difference between the last two positions was only 0.12 points.

That kind of compression matters. It means seemingly small edge cases—navigation dropouts near the finish, failed three-point turns, blocked parking exits, or accident vehicles straddling solid lane markings—can materially alter the standings. D1EV says these disputes led to supplementary provisions in version 8.5 of the competition rules, centered on one principle: external objective conditions should not be used as reasons to waive penalties. Teams can still publicly annotate questionable disengagements and apply for post-run retesting.

This is an important marker for the Chinese EV market. Advanced driver assistance systems are now good enough that the ranking battle is increasingly decided by the messy, low-speed, high-ambiguity situations that define real urban driving.

Aion Emerges as the Biggest Threat to Exeed

The headline story in the final rankings is GAC Aion’s sudden rise. Aion entered the official final leaderboard for the first time and immediately took second place with an average score of 96.86, only 0.28 points behind the leader.

Exeed, using a Bosch + WeRide solution, remains first at 97.14, but its lead has been shrinking. Earlier, its advantage had been 0.35 points. The latest rolling window, covering competition rounds 26-31 from April to September 2026, dropped older Jinhua results and added Shenyang, tightening the table further.

Final leaderboard snapshot

Aion’s numbers are especially impressive because its trend line is moving up quickly:

• Guiyang final: 91.96
• Shenyang final: 101.76
• Preliminary average: 103.43
• Shenyang preliminary score: 105.5, the highest in the field

Just as important, Aion sharply reduced penalties. In Guiyang it lost 22 points, including 15 points for “ghost behavior”. In Shenyang, total deductions fell to 11 points, with zero penalties in all-scenario capability, avoidance, ghost behavior, and accident modules.

Exeed’s story is more complicated. Its historical scores remain strong—105.5 in Wuhu, 102.81 in Hefei, 94.04 in Kunming—but Tianjin dropped to 86.21, hurt by 15 points in all-scenario deductions and 12 in efficiency. More concerning is the preliminary-stage slide: 105.5, 101.88, 99.5, 96.5, then 85.64 across five stations. D1EV notes that the EXEEDOS 2.8.0 version has gone roughly half a year without an update, giving rivals time to catch up.

Huawei’s Mixed Picture: Aito Improves, Avatr Stalls

Huawei’s intelligent driving stack continues to show both high promise and persistent blind spots.

Aito entered the leaderboard for the first time and landed directly in fourth place with an average final score of 89.20. In Shenyang, it advanced from a 91.5 preliminary score to 94.57 in the final, good enough for third place in that event. That was 10.75 points higher than its Kunming final result of 83.82.

The biggest improvement came in efficiency:

• Kunming deductions: 28 points total, including 20 for all-scenario ability and 8 for efficiency
• Shenyang deductions: 19 points total, including 15 for all-scenario ability, 2 for violations, and just 2 for efficiency

However, D1EV highlights a familiar Huawei ADS issue: performance in park or campus waypoints appears highly dependent on prior map coverage. If the area has been mapped, the system is more likely to activate smoothly; if not, it often fails to engage. That remains one of the key obstacles between Huawei’s strong highway/city NOA promise and true all-scenario consistency.

Avatr, also using Huawei technology, tells a less encouraging story. Its average score stayed at 80.63, but it slipped from fourth to fifth as newer entrants moved ahead. More worrying, it has missed the final in three straight rounds: Kunming, Guiyang, and Shenyang.

XPeng and Other Challengers: Progress, But Not Clean Progress

XPeng, listed here with its vision-based stack, participated in six finals and posted an average score of 78.22, up only 0.22 points from the previous period. That sounds modest, but the trend in Shenyang was at least a partial recovery:

• Shenyang final: 80.07
• Improvement of 9.73 points versus Guiyang
• Ended a run of three straight sub-80 scores

There was measurable progress in efficiency, where XPeng cut deductions from a full 30 points in Guiyang to 14 in Shenyang. It also recorded its second zero-deduction all-scenario result in the past six months.

Still, new weaknesses emerged:

• Avoidance deductions: 5 points
• Violation deductions: 6 points
• Ghost behavior: 5 points, the first such penalty in six months

Elsewhere, Exeed’s Horizon-based solution remains competitive but too inefficient in dense real-world scenarios, while Chery Fulwin continues to look volatile—capable of a 100.87-point win in Kunming, but also vulnerable to severe deduction-heavy performances.

Why These Rankings Matter More Than a Typical Test Event

D1EV’s competition format is increasingly becoming a useful proxy for the real state of China’s assisted-driving race. Unlike controlled proving-ground demonstrations, these rankings expose systems to dense urban complexity, imperfect mapping, temporary obstructions, and route-planning ambiguity.

Key lessons from the latest results include:

• Consistency is now more valuable than peak scores
• Low-speed urban edge cases are becoming the main battleground
• Software update cadence matters as much as hardware capability
• Map dependence remains a major differentiator among Chinese ADAS stacks

For Chinese EV brands like GAC Aion, Aito, Avatr, XPeng, and Exeed, this is no longer just a branding exercise. It is becoming a measurable test of whether their autonomous driving claims can survive the unpredictability of Chinese cities.

NavInfo Wins Beijing Smart-Road Project Worth RMB 100.26 Million

While automakers battle for in-car intelligence, another piece of the puzzle is being built outside the vehicle.

NavInfo has won a Beijing “Dual-Intelligence” City 4.0 project in Fangshan District, with a contract value of about RMB 100.26 million. The project covers 190 square kilometers and 398 intersections, focusing on roadside perception equipment and data transmission to multiple government and municipal platforms.

This is notable for two reasons.

First, it is NavInfo’s first direct roadside perception equipment contract as the lead bidder in Beijing’s Smart City 4.0 phase. Second, the project shifts the emphasis from pure vehicle-road autonomous driving applications toward urban governance and traffic management.

NavInfo Beijing project at a glance

NavInfo says the win validates both its foundational data capabilities and roadside perception delivery capacity. The company had previously participated in Beijing’s Vehicle-Road-Cloud 3.0 construction in Tongzhou as part of a consortium, covering 660 square kilometers of high-definition map production and applications, combined with roadside data for map change detection and updates.

That progression—from Tongzhou’s “spatiotemporal base layer” to Fangshan’s “perception delivery”—suggests that China’s smart mobility ecosystem is moving beyond pilot projects. The market is shifting toward operational value extraction, where HD maps, roadside sensors, cloud connectivity, and traffic management systems are integrated into a single stack.

Pony.ai Expands in the UK as Chinese Robotaxi Globalization Accelerates

The third major story is Pony.ai’s accelerating push into Europe.

On October 14, Pony.ai announced that its seventh-generation Robotaxi had officially landed in Manchester, where vehicle testing is already underway. What makes this striking is the speed: the Manchester announcement came less than one week after Pony.ai confirmed its London project on October 8.

Manchester is the UK’s second-largest urban agglomeration and a major transport and economic hub in northern England. Compared with London, it offers a more representative mid-sized European city profile, with:

• Multi-center urban structure
• Dense short-distance travel demand
• Strong suburban connector traffic
• Mixed commuting patterns

That makes it an ideal test bed for validating autonomous driving in a broader set of European conditions.

Pony.ai is not expanding alone. Its model combines:

• Pony.ai: autonomous driving technology
• Verne: local fleet operations
• Uber: passenger access via its global mobility network

This framework was already tested in Zagreb, Croatia, where Pony.ai launched what the source describes as Europe’s first commercial Robotaxi service integrated with Uber and Verne. After commercial service began in April, the company moved to fully driverless passenger testing within five months.

The bigger strategic target is ambitious: Pony.ai is working toward a European deployment plan spanning five cities and more than 2,000 Robotaxis, while also expanding in the UAE, Singapore, and South Korea.

Global Implications

Taken together, these three stories show how the Chinese EV industry is evolving into a broader intelligent mobility industry.

1. Vehicle intelligence is entering a precision phase

The D1EV rankings show that the top Chinese assisted-driving systems are no longer separated by huge performance gaps. Leaders and challengers are now fighting over fractions of a point, with victory hinging on unusual urban edge cases rather than basic lane-centering or highway navigation.

2. Infrastructure is becoming a competitive moat

NavInfo’s Fangshan win highlights a major trend: smart roads, HD maps, roadside sensing, and vehicle-road-cloud integration are becoming critical assets. In China, autonomous driving will likely be shaped not only by what cars can do independently, but by how effectively they plug into city-scale digital infrastructure.

3. Chinese autonomous driving companies are exporting operating models, not just software

Pony.ai’s UK moves suggest that Chinese firms are learning how to package autonomy as a repeatable international business model. The combination of technology provider, local operator, and global platform partner may become a preferred route for Robotaxi expansion in markets with complex regulatory and operational requirements.

Why This Matters for the Chinese EV Market

For global observers, these developments reinforce a point that is sometimes missed outside China: the country’s EV race is no longer just about battery scale, low-cost manufacturing, or export growth. It is increasingly about who can master three interconnected layers:

• In-car intelligence through ADAS and urban NOA
• City-side intelligence through roadside sensing and cloud infrastructure
• Mobility services intelligence through Robotaxi fleet deployment and operations

Brands such as GAC Aion, Aito, Avatr, XPeng, and Exeed are competing in the first layer. NavInfo and similar suppliers are building the second. Pony.ai is proving the third can travel overseas.

What Comes Next

The next few months will be important on all three fronts.

In assisted driving, the key question is whether Aion can sustain its momentum and overtake Exeed as older high scores roll out of the six-month window. Huawei-backed systems, especially Aito, will also be watched closely for improvements in park and campus scenarios where map dependence remains a weakness.

In infrastructure, Beijing’s Smart City 4.0 rollout will be a test of whether vehicle-road-cloud projects can deliver measurable traffic-management and governance value rather than just serving as technology showcases.

For Robotaxis, Pony.ai’s progress in Manchester and London will indicate whether Chinese autonomous driving players can move from isolated pilots to scalable multi-city international operations.

The broader conclusion is clear: China’s mobility leaders are now competing on roads, in clouds, and across borders. The EV story is still about electrification—but increasingly, the winners will be decided by autonomy, data, and deployment discipline.