MOMENTA STRATEGY

Momenta pursues a two-leg strategy of series production (Mpilot) and full autonomy (MSD). Mpilot, for series production vehicles, generates massive amounts of data. The data is fed into closed-loop automation process to improve data-driven deep learning algorithms, which improves MSD technology. Technology advancements from MSD, in return, are implemented in updates of series production (Mpilot). Data, data-driven algorithms, and closed-loop automation between Mpilot and MSD are the key to achieving autonomous driving at scale.

Date Flow
Technology
Advancement
Scroll

CLOSED LOOP AUTOMATION

Closed loop automation is a systematic approach to achieve rapid iteration of autonomous driving algorithms. It connects the entire research and development iteration process, and forms closed feedback loop. The automated feedback loop reduces repetitive work, focuses on core values, and thus optimizes the product development cycle.

Deep Learning Foundations

Big Computation

Large-Scale Deep Learning
GPU Training System

4000GPUs Large-Scale DL Training Cluster “ROCS” Momenta designed distributed DL training system

4000GPUs Large-Scale DL Training Cluster “ROCS”
Momenta designed distributed
DL training system

Big Data

2Billion kilometers yearlyData Collection Capability

Current Collection Capability 22,000+ crowdsourcing vehicles >1 billion labelled objects yearly

Current collection capability
22,000+ crowdsourcing vehicles
>1 billion labelled objects yearly

Model Optimization

Model size compression
with retained accuracy

100x Speed boost

100x Speed boost

Embedded device friendly

10-100x speed boost

Core Technology

Perception of Pedestrians,
Vehicles, and Roads.

Perception of Pedestrians,
Vehicles, and Roads.

HD Map

HD Map for autonomous driving entails features of precision, freshness and scalability. 3D positioning of the surrounding environment is reconstructed by extracting semantic information from 2D images and fusing with GPS and IMU data. “Live” map which powers various levels of autonomous driving is automatically generated from crowdsourcing vehicles with low-cost and mass-produced sensors. HD Map is used by different autonomous driving software modules such as localization, planning and control.

HD Map

HD Map for autonomous driving entails features of precision, freshness and scalability. 3D positioning of the surrounding environment is reconstructed by extracting semantic information from 2D images and fusing with GPS and IMU data. “Live” map which powers various levels of autonomous driving is automatically generated from crowdsourcing vehicles with low-cost and mass-produced sensors. HD Map is used by different autonomous driving software modules such as localization, planning and control.

Data-Driven Path Planning

We aim to build a smart driver with billions of miles of driving experience using data-driven approach. Crowdsourcing allows us to obtain billions of actual driving trajectories memorized in the localization module of HD Map. We perform autonomous driving planning by mapping from environment perception data to driving trajectories in HD Map. This provides us a unique and elegant framework to solve corner cases by adding corresponding data rather than adding rules.

Data-Driven
Path Planning

We aim to build a smart driver with billions of miles of driving experience using data-driven approach. Crowdsourcing allows us to obtain billions of actual driving trajectories memorized in the localization module of HD Map. We perform autonomous driving planning by mapping from environment perception data to driving trajectories in HD Map. This provides us a unique and elegant framework to solve corner cases by adding corresponding data rather than adding rules.

Join the future ride with us @Momenta!

Beijing
3rd Floor, Block C, Dongsheng Building, 8 Zhongguancun East Road, Haidian District,
Beijing, China
+86(010)82526609
contact@momenta.ai
Suzhou
21st Floor, Trirun Soho, Nantiancheng Rd, Xiangcheng District
Suzhou, China
+86(512)66182558
contact@momenta.ai
Stuttgart
Momenta
Rotebühlplatz 23
70178 Stuttgart, Germany
+86(010)82526609
contact@momenta.ai