We less doubt the company’s academy capability. A deep relationship with OEMs can be a two-edged sword, however, as well as commercialization too earlier.
March 25, 2019/ EqualOcean/ - China-based self-driving company Momenta (初速度) plans to launch fully autonomous driving solutions in the third quarter of 2019, following the company’s an autonomous driving solution targeting expressway scenarios in the first quarter and an auto parking solution in the second quarter, according to the company’s CEO CAO Xuedong (曹旭东) at a conference that celebrated the inauguration of its Suzhou headquarters, according to iyiou.com.
CAO, is a former researcher in Microsoft Research Asia (MSRA) and a former Executive R&D Director of SenseTime, the most valuable AI startup in China backed by Alibaba.
The company’s fresh new building located at Yangcheng Eco-district, an important intersection of Road & Belt (R&B) and Yangtze River Economic Belt (YREB).
As we have reported before, the company abandoned its former headquarter Beijing and chose to move to Suzhou. 60% of projects were cut as well, according to people familiar with the matter (see more in this article).
The company is eager to commercialize and now let’s see how they managed it. The company plans to unveil its autonomous driving solutions targeting expressway scenarios in Q1 and auto parking solution in Q2. In addition, an autonomous driving solution designed for urban scenarios will be released in the first quarter of 2020. Moment is able to mass produce all solutions for expressway, parking, and urban scenarios, according to CAO. CAO also emphasized that those solutions are based on mass-produced sensors.
“The key to launch before-assembly mass production products is to productize autonomous driving technology by combining deep learning with mass produced sensors and computing processors, ” CAO claimed reported by one of Momenta’s backers NIO Capital.
Momenta exhibited its structured driving solution Mpilot, an automotive grade, and mass produced sensor solutions which utilize camera, high-definition map, and positioning technology in a sensor fusion way. The solution can operate perfectly on the rainy day, traffic jam in the dusk, and night.
The company is cooperating with several OEMs and Tier 1 manufacturers from America, EU, Japan, and China to produce in before-assembly market.
As for self-driving technology improvement, CAO claimed that it is necessary for the company to “walk in two legs”. On the one hand, it obtains a huge amount of sensor and mapping data by mass producing autonomous driving solutions and then utilizes these data in road tests for fully autonomous driving solutions. On the other hand, the road test results should be reported to the team that is responsible for mass production, so as to keep optimizing algorithm and form a technically closed loop, according to Gasgoo.
Based on the concept, Momenta has built a series of basic architectures for big data, big computing, modeling, and algorithms. The company has built a computing platform with more than 4,000 GPUs.
In HD map domain, Momenta has realized an accuracy in mapping and positioning within 10 cm with sensors and devices such as monocular camera, automotive-grade IMU (Inertial measurement unit), GNSS (global navigation satellite system) and chips.
We less doubt the company’s academy capability. For instance, Momenta has joined in building the widely used Squeeze-and-Excitation Networks and the related paper ranked first in 2018 Google Scholar Search Ranking. A deep relationship with OEMs can be a two-edged sword, however, as well as commercialization too earlier (see more in this article).