Topics Price war Tariffs Solid state ADAS Software Sales data Supply chain Battery recycling

Home /Data

Data 4 min read

Aion closes gap on Exeed in China's ADAS league table as rules tighten

Aion's two straight wins put it second in China's intelligent-driving competition standings, just 0.28 points behind leader Exeed, while rule changes address edge cases like failed NOA activation and rerouted three-point turns.

China's intelligent-driving competition, organised by the outlet D1EV (第一电动), published its September standings, and the margins at the top have narrowed to a razor-thin 0.28 points. GAC Aion, which won the last two rounds in Guiyang and Shenyang, entered the final standings for the first time and jumped straight to second place, trailing leader Exeed by just 0.28 points. The gap between the last two ranked brands is only 0.12 points, according to the half-year rolling table.

The tight scores prompted the organisers to issue a supplement to version 8.5 of the competition rules after internal disputes over scoring in Shenyang. One argument concerned a vehicle that could not activate its NOA (navigate-on-autopilot, a form of urban advanced driver assistance) near the finish because GPS positioning failed - should that count as a penalty? Another involved a three-point turn at checkpoint 7 on Bashan Road, where none of the 11 finalists completed the manoeuvre. Aito and Li Auto instead let the navigation system reroute around the spot. The new rule states that external objective conditions cannot be used as grounds for reducing penalties, but drivers may request a post-race review and have disputed interventions noted publicly.

Exeed's lead shrinks

Exeed, using Bosch and WeRide technology, has topped the final standings for seven consecutive periods, but its average score fell from 100.12 to 97.14 as the 112.05-point Jinan result rolled out of the six-month window. Its four finals scores were strong - Wuhu 105.5, Hefei 102.81, Kunming 94.04 - but Tianjin dropped to 86.21. More worrying is the preliminary-round trend: 105.5 in Hefei, 101.88 in Tianjin, 99.5 in Kunming, 96.5 in Guiyang and 85.64 in Shenyang, four consecutive declines. Its EXEED OS 2.8.0 software has not been updated for six months, giving rivals a window. If Exeed misses the finals again in October, its average could fall to 94.35, below Aion's current 96.86.

Aion, using WeRide technology, has only two finals results but both were wins: 91.96 in Guiyang and 101.76 in Shenyang, with total deductions shrinking from 22 points to 11. In Shenyang it recorded zero deductions across the all-scenario, avoidance, ghost-behaviour and accident modules. Its preliminary average of 103.43 is the highest of any brand. The caveat is the small sample size - two events - and whether it can sustain this form across different cities.

Exeed (Horizon Robotics) dropped from second to third with an average of 92.58 after missing three finals and returning in Shenyang with 88.87, fifth on the day. Its efficiency module remains the weak point, with deductions rising from 4 to 11 to 15 points across Wuhu, Hefei and Shenyang. The Horizon V2.0 software is a clear improvement over V1.6.7, but the car still hesitates to overtake and can get stuck even when there is space to pass.

Huawei family reshuffle

Aito (Huawei) entered the final standings for the first time, taking fourth with an average of 89.20 from two finals, displacing Avatr as the highest-ranked Huawei-system brand. In Shenyang it scored 94.57, third on the day, up 10.75 points from its Kunming result of 83.82. Its efficiency deduction fell to just 2 points, the lowest for a Huawei-system brand in six months. But it still lost 15 points in the all-scenario module in Shenyang, all at a campus waypoint where the system could not activate. In the preliminary round the same waypoint caused no problem. Huawei's ADS has long struggled at unscanned campus waypoints, and the gap to the top remains there; a new version is said to address it, with the Changsha round as the test.

Avatr (Huawei) held its average at 80.63 but slipped from fourth to fifth because Aito and Aion entered the standings. It has missed the last three finals. Its all-scenario capability is the biggest deduction source: 15 points each in Wuhu and Hefei, 30 in Tianjin. Its preliminary scores slid from 99.5 in Tianjin to 73.5 in Guiyang before recovering to 80.5 in Shenyang. Chery Fengyun (Horizon Robotics) stayed at 79.07 but dropped from fifth to sixth after failing to qualify for the Shenyang final with a preliminary score of 76.69. Its Kunming win of 100.87 still supports its average, but Guiyang cost it 49 points, including 28 for efficiency and 10 for a traffic incident.

Xpeng, using a vision-based system, improved its average slightly to 78.22 but fell from sixth to seventh as new brands entered. Its Shenyang result of 80.07 ended a run of three consecutive finals below 80. The efficiency module improved markedly, from a full 30-point deduction in Guiyang to 14 in Shenyang, and it recorded zero all-scenario deductions for the second time. But new problems appeared: 5 points for avoidance, 6 for traffic violations and its first ghost-behaviour deduction in six months, worth 5 points.