Backstage. Win the Client

Begin with the end.

Everything dates back to February 2019 when Maxwell and a few partners founded DeepRoute.ai and envisioned autonomous driving to be accessible and affordable for everyone. However, the challenge remained in achieving it with a giant computing platform that consumes significant energy and the limitation of Robotaxis and expensive mechanical LiDAR.

?

Strategic Thinking

Unlike others who keep expanding their Robotaxi fleet, DeepRoute thinks that working with OEMs with a lower-cost, auto-grade solution is the best approach. Though a small team, it runs swiftly and solves questions creatively. An inference engine that speeds up neural network computing was debuted in January 2021, enabling autonomous driving algorithms to run on an energy-efficient computing platform.

In December, DeepRoute launched Driver 2.0, which was the first in the industry to set the lowest BOM price at that time—US$10K. Mechanical LiDARs are out, and Solid-state LiDAR is in, pushing technology progression to faster and faster speeds. Eventually, the Driver 3.0 Mapfree solution was released in March 2023 with only 1 LiDAR and 7 cameras, priced at $2,000 for the BOM.

1 hour. 3 weeks. Days and months.

"We put together all the resources and pulled all-nighters to finish retrofitting the client's car within 3 weeks, encompassing both drive-by-wire and software integration. We were racing against time. Eventually, the client was touched not only by our solidarity and devotion but also by our technological advantage when they experienced firsthand how their existing car would look and drive with our smart driving functions," said Tian, Head of Business Development. It was an opportunity too precious to be wasted after DeepRoute earned the client's trust by retrofitting their car after a morning rush hour Robotaxi demo in the city center with zero intervention.

?

Everything seems promising, but a few questions remain pending before winning the bid officially: How can we successfully work with OEMs on the mass production of smart driving cars with L2 and above functions, considering it is a Herculean task? How do we develop all the functions from scratch and ensure they pass the safety tests? Additionally, how can we deliver the same Robotaxi experience with the mapfree solution? Lastly, what is the best hardware configuration for the client's car model?

?

The reorganization

"I almost cried when the team developed this function within weeks. There's another supplier who develops independent modular functions, and the client was making a comparison. If we don't do this function well, we will lose," said Tian.

In contrast to previous L4 Robotaxi demos, all the modular functions such as AEB, APA, and TJA now have precise product definitions and boundaries. DeepRoute swiftly reorganized to meet the mass production project's requirements, moving away from the previous algorithm-centric team structure. It has been a challenging yet rewarding process of advancement. Forward development was executed with product definition, function development, and mass production quality assurance all prepared.

The proposed L2+ solution, featuring 1 LiDAR, 1 SoC, and other configurations, costs only half as much as competitors with a similar level of competency. In the third quarter, a successful mapfree demo was showcased to the client on the highway, followed by urban scenarios one month later.

?

With tremendous effort, the bid was secured at the end of the year.

?

Talents, young and senior

A key part of the winning strategy is the distinctive mapfree technology. Since universal standards worldwide are yet to be established, DeepRoute has assembled a team of cross-disciplinary talents who think outside the box and tackle these undefined problems. With a mix of knowledge in SLAM, SD map, deep learning, and more, they continually inspire and innovate brand new approaches. "It is my honor to be involved in this project. It's super rewarding to see my lines of code executing functions and bringing safety to consumers when they're driving on the roads," said Mo, who graduated just two years ago.

Veteran engineers, well-versed in the ins and outs of the automotive industry, significantly enrich the company's overall competency, ensuring functional safety, testing, and successful delivery to the client.

Winning the client and reorganizing is merely the first step. The key lies in rolling out the mapfree solution from highways to urban scenarios and ensuring timely deliverables. In the next chapter, we will delve into how the team achieved the mapfree solution and share their personal stories during this journey. To be continued.