Haomo.ai unveils affordable map-free NOH solutions at HAOMO AI DAY

Beijing (Gasgoo)- On October 11, 2023, Great Wall Motor-backed autonomous driving solution developer Haomo.ai held the ninth HAOMO AI DAY event, at which the company introduced its latest product offerings and technological achievements in the smart driving field.

Photo credit: Haomo.ai

To meet the dynamic intelligent driving market, Haomo.ai launched three "extremely cost-effective" NOH (Navigation on Highway) solutions for various vehicle production requirements, covering premium, mainstream, and low-end vehicle segments. The intelligent assisted driving products - HP170, HP370, and HP570, are scheduled to be deployed onto vehicles in 2023 and 2024.

The HP170 is an ultra-affordable high-speed NOH solution, priced at approximately 3,000 yuan. With 5 TOPS of computational power, it features one front-view camera, four fisheye cameras, two rear corner radars, and twelve ultrasonic radars. Designed for high-speed and urban fast-road NOH, it excels in short-range memory parking and is certified with a high safety standard, E-NCAP 5-star AEB.

Photo credit: Haomo.ai

Priced around 5,000 yuan, the HP370 offers city driving and parking with memory capabilities. Boasting 32 TOPS of computing power, it comes with two front-view cameras, two side-view cameras, one rear-view camera, four fisheye cameras, one front radar, two rear corner radars, and twelve ultrasonic radars. It caters to high-speed, city driving and offers memory-based parking, intelligent obstacle avoidance, and urban driving functions.

The HP570, available for about 8,000 yuan, offers a comprehensive urban NOH solution. It provides the choice of computational power with 72 or 100 TOPS chips and includes two front-view cameras, four side-view cameras, one rear-view camera, four fisheye cameras, one front radar, twelve ultrasonic radars, and the option to add a LiDAR. Designed for urban NOH, it covers all-scenario parking assistance, intelligent obstacle avoidance, and multi-layer memory parking, all at a highly competitive price point.

At the HAOMO AI DAY, the company presented Haomo.ai's latest accomplishments in its four major campaigns for 2023.

Photo credit: Haomo.ai

In terms of the installation volume of smart driving products, Haomo.ai has solidified its position as China's leading producer of autonomous driving systems. Its HPilot, an advanced driver-assistance system, is now integrated into over 20 vehicle models. Users have collectively traveled over 87 million kilometers with the assistance of HPilot. Notable models among the latest vehicles equipped with Haomo.ai's HPilot are the Shanhai Poer HEV pickup truck, and the new WEY Mocca Hi-4S.

HP550 in  WEY Blue Mountain; photo credit: Haomo.ai

Regarding urban NOH deployment, Haomo.ai's HP550, previously known as HPilot 3.0, is set to commence mass production and deployed onto the WEY Blue Mountain model in the first quarter of 2024. 

Xiaomotuo 3.0; photo credit: Haomo.ai

As to terminal logistics autonomous delivery development, Haomo.ai’s unmanned goods delivery vehicle, the Xiaomotuo 3.0, has delivered over 220,000 orders and is set to achieve profitability in the fourth quarter of 2023 in supermarket fulfillment and delivery scenarios. From a business perspective, Haomo.ai possesses the capability to become the first company globally to have a profitable large-scale L4 autonomous driving business. The Xiaomotuo 3.0 is priced at 89,999 yuan and is the world's first medium-sized autonomous delivery vehicle under 90,000 yuan. It caters to a wide range of scenarios, including logistics, supermarkets, and retail. 

At the event, Haomo.ai also provided an in-depth overview of the advancements achieved by its DriveGPT autonomous driving large model after approximately 200 days since its release.

Photo credit: Haomo.ai

As of October 2023, DriveGPT has meticulously curated an extensive dataset, consisting of over 10 billion frames of internet images and 4.8 million segments of 4D Clips data, which includes human driving behaviors. This achievement has attracted 17 ecosystem partners, resulting in a 90% efficiency improvement for these partners.

Photo credit: Haomo.ai

Moreover, DriveGPT has significantly improved its general perception capabilities. This is achieved by introducing a multi-modal large model that integrates text, images, and video data to acquire the ability to recognize a wide range of objects. Furthermore, by integrating NeRF (Neural Radiance Fields) technology, DriveGPT has achieved enhanced 4D spatial reconstruction capabilities, allowing for comprehensive modeling of three-dimensional space and temporal sequences.

In addition, leveraging large language models, DriveGPT has successfully integrated a wealth of global knowledge into its driving strategies.

Haomo.ai showcased three significant test results that demonstrate DriveGPT's contributions to in-vehicle capabilities at the event as well.

Photo credit: Haomo.ai

First of all, Haomo.ai showcased successful results in purely visual autonomous parking tests. Utilizing visual perception models, with fisheye cameras to identify boundaries of objects such as walls, pillars, and other vehicles, the system achieved a full 360-degree dynamic perception within a 15-meter range, with a measurement accuracy of 30cm and even higher accuracy within 2 meters (less than 10cm). This level of perception accuracy opens the possibility of replacing traditional ultrasonic sensors with visual sensors, potentially reducing the overall cost of autonomous driving solutions.

Photo credit: Haomo.ai

Secondly, DriveGPT has evolved its perception capabilities from recognizing only a limited number of obstacles and lane markings to identifying all elements in traffic scenarios. This includes various traffic signs, road markings, and even features like manhole covers. The availability of high-quality, fully annotated data for these scenarios has greatly improved the effectiveness of DriveGPT's in-vehicle perception model, further accelerating the deployment of urban NOH systems.

Photo credit: Haomo.ai

Finally, Haomo.ai's tests on DriveGPT's urban NOH showcased its ability to effectively detect small obstacles at distances up to 50 meters, typically with a height of approximately 35cm, even at speeds of up to 70 kilometers per hour. This includes detecting and successfully avoiding or stopping for small, moving obstacles such as animals on the road, offering a robust safety feature in urban traffic environments.