With a funding amount of 128 million USD in May this year, Roadstar.ai achieved probably the highest single-round financing amount, among all the autonomous driving solution startups of China.
Compared with 2017, entrepreneurship environment of this field is silently changed. Startups’ disparity is coming, those with better technology strength and achievements in the test will attract much more attention of the venture capital as well as traditional auto market players, like manufacturers and T1 suppliers, etc.
Roadstar.ai’s record-breaking (which was previously kept by Pony.ai by USD 112M) amount in its Series A Funding somehow reflected the recognition of technical ability from the capitals, which also helped Roadstar.ai to join the top startups club of the self-driving solution. In terms of valuation, Roadstar.ai is now ranked the third with USD 600 million, right after Pony.ai and Momenta (both with a valuation around USD 1 billion).
China's self-driving tech industry has entered a new competitive stage of "productization"
Maybe inspired by Waymo, or maybe the great minds thought alike, Chinese self-driving startups also reached a consensus on the route of future productization – robotaxi. However, the robotaxi is a relatively middle-term plan, which might be realized in at least several years, and will be highly determined by more than technology, like policies and market acceptance.
Although mature productions are not ready yet, self-driving tech startups started to show their “muscles” to the market, especially those VCs and traditional auto manufacturers, as well as governments, to gain confidences, and the most important, funds to survive and to develop.
All the three key competitors of self-driving technology had launched their “product-ready” solutions, while with different core competencies:
Pony.ai launched PonyAlpha and was invited to the 2018 WAIC to serve with its autonomous fleet. T3 license, the autonomous fleet in Guangzhou and the genius founding team are the highlights of this company;
Momenta also made its own autonomous vehicles of different levels, obtained Shanghai City’s road test license, with a story of crowdsourcing data collection mode, which was claimed more efficient with a much better return of investment.
Roadstar.ai focused on storytelling of its technology of Multi-sensor Information Fusion (MSIF) preprocessing, which includes two self-developed technologies: HeteroSync, which makes sensors of different brands, types and quantities compatible, and achieves precise synchronization of time and space; DeepFusion, which integrates the 8-dimensional data of lidar, camera and millimeter-wave radar at the original data level. Roadstar.ai’s preprocessing MISF technology could produce a truly robust, efficient and safe autonomous driving solution, it was even said that 4 thousand data could achieve a result which might need 150 thousand data by using a traditional algorithm.
Screenshot of Roadstar.ai’s road test in Silicon Valley
As to the productization, enhanced by its core algorithm, Roadstar.ai launched its Level 4 autonomous solution “Aries.锐” in the first half of this year, and brought its whole new hardware solution on the press conference held in Wuzhen on November 7, during the 5th World Internet Conference (WIC).
Roadstar.ai served WIC in Wuzhen with “Leo”
Roadstar.ai’s new generation self-driving solutions "Leo.灵", is a Level 4 autonomous driving solution upgraded from the previous edition “Aries.锐”, and is said could better manage the complex road conditions in Chinese cities.
With Leo, Roadstar.ai’s autonomous fleet served the WIC guests in Wuzhen, before which, “the team of Roadstar.ai only spent one week to prepare, which was said cost Baidu almost half a year two years ago here” said HENG Liang in a recent telephone interview with EqualOcean.
In addition, roadstar.ai has also unveiled a new map technology in the press conference, named HD RealityMaps, which could provide a colorized 3D information to the autonomous driving system, and could better meet the “needs” of self-driving vehicles for the map.
Better than Waymo? Must be kidding
Roadstar.ai once claimed in an interview published on its official website before, that “Google's algorithms are relatively mature, however, it relies too much on lidars which are now too expensive, Roadstar.ai aimed to provide better performance than Waymo, with a lower cost” (by MSIF mentioned above).
It is indeed an ambitious objective and should be encouraged. Meanwhile, we do need to face the truth that Roadstar.ai in June 2018, claimed to accomplish one-disengagement-per-500-kilometers road test in California, successfully achieved ONLY 1/18 of Waymo did one year before (an average of 9009.5 km/D), according to the Autonomous Vehicle Disengagement Reports 2017 published by the Department of Motor Vehicles, State of California.
Waymo’s all reportable disengagements
On the other hand, technical route of Multi-sensor Information Fusion is actually not Roadstar.ai’s differentiation, a bunch of self-driving companies around the world, including Waymo, also uses similar technologies.
Instead of talking about chasing Waymo, perhaps Roadstar.ai should rather pay more attention to the real business, after all, it has no deep pockets like the Waymo and really needs to survive first.
Besides Level 4 autonomous solutions, Roadstar.ai also planned to provide a level 2/3 cost-effective design solution to help automotive partners add technology values and achieve a safer driving experience.
By whom will Roadstar.ai’s solution and even its technical consultation services be bought, remains the most serious question.