Gasgoo Awards 2023 applicant: MAXIPILOT®2.0 High Performance Intelligent Driving System -Deploys BEV Transformer Network, Supports Driving-parking Integration Solution from Shanghai MAXIEYE Automotive Technology Co., Ltd.

Product: MAXIPILOT®2.0 High Performance Intelligent Driving System -Deploys BEV Transformer Network, Supports Driving-parking Integration Solution 

Description

MAXIPILOT®2.0 is a high-performance intelligent driving solution based on the BEV transformer network and the 8M visual perception platform. It supports 1V1R/1V2R/1VnR+4fisheye sensor configuration and could realize L0-L2+ intelligent driving control functions, including high-speed NOA. Both perception accuracy and system performance of MAXIPILOT®2.0 have been significantly improved based on BEV perspective and end-to-end spatial perception and timing fusion, which improves the safety, reliability, intelligence and comfort of users’ driving experience. MAXIPILOT®2.0 can also be configured as a domain controller platform to develop into an integrated driving-parking system solution with “system cost advantage”. MAXIPILOT®2.0 supports the deployment of OTA online upgrade technology and shadowing mode, and has realized innovation closed loop of data-driven technology, product and service. Supported functions: LDW/LCC/iACC/AEB/TJA/AHBC/TSR/TLR/BSD/DOW/FCTA(B)/ESA/ILC/ALC/NOM/APA, etc.

Unique advantages

1. With clearly defined product functions, the product could eradicate all pain points and users’ rigid demands to match cost and experience reasonably. On the basis of safety and comfort, MAXIPILOT®2.0 supports intelligent functions such as high-speed NOA, memory driving, memory parking pilot, GOD (general object detection), urban area intersection traffic assistance enhancement, etc. The platform supports the upgrading of all fields, ranging from algorithm, product form to full-dimensional experience.  2. Self-developed of intelligent driving core technology with full-stack. MAXIEYE has independently mastered the full-stack core technology of intelligent driving from perception to planning control. The image-level original truth value and generalization scene understanding brought by full-stack self- developed will greatly reduce the error superposition caused by cross-platform research and development between perception and control. It will enable customers to achieve customized product development, and will bring consumers more extreme product experience.  3. Leading AI algorithms. By applying BEV technology architecture to MAXI-Net, MAXIEYE proposed the end-to-end intelligent driving deep learning perception method, which makes transformation from image space perception to BEV space perception, integrates time sequence information, supports multi-modality across cameras and other sensors, which greatly improves perception accuracy, better solves target occlusion and optimizes feature fusion. It also simplifies and reduces the weight of the technical system perceiving integrated control, and makes the whole automatic driving system simpler and more efficient.  4. Based on FPP technology, the system effectively reduces taking-over rate and improves system fluency. The system can make full use of various information of the surrounding environment, and make comprehensive judgement with priority of navigation, lane lines, traffic flow, railings, guardrails, roadside and other information. It keeps the vehicle driving smoothly in the scene of missing high-precision map or missing navigation.  5. Strong ability of embedded software development. MAXIEYE can maximize the use of chip computing power resources, and can achieve fast migration between different computing power platform chip, so as to support costs reduction by design without performance sacrifice.  6. Flexible and customized solutions. Based on the full stack capability, MAXIPILOT®2.0 offers users different driving style options that match different driving strategies, from efficiency first mode to comfort mode, etc.   7. The product supports the high integration of driving and parking functions. The reuse and mutual complement of driving and parking sensors bring more extreme product experience to users. Based on engineering mass production accumulation and large-scale data verification, the system can properly handle different complex scenarios on roads (such as: tunnels, curves, complex lane lines, extreme CUT-IN, etc.); in terms of parking, the advantages of autonomous long-range forward vision can facilitate the parking system’s positioning function and improve agility (parking at a faster speed).  8. Data driven. MAXIEYE has completed the data accumulation of 200 million km real road conditions through the self-developed data closed-loop system, and established the data intelligence system based on AI algorithm. On the basis of the designs above, MAXIEYE has been fitted with the deep learning algorithm and scene landing, which are relatively advanced in the industry at present. Based on OTA online transmission technology, MAXIEYE further expands road condition database resources for customers to accelerate product update iteration driven by data, and provide users with after-sales support such as OTA software upgrade.

Application

Highway, Urban Road

Prospect

The era of high-level intelligent driving has come. For the next generation of high-level solutions, “slogans” outweigh “sales volume” in the industry. As of 2022, the penetration rate of driving-parking integration on the domestic market was still less than 5%, which means a huge market potential. According to statistics, the total mileage of highways in China is 5.35 million km, including 177,000 km expressways, accounting for only 3% of the total mileage. However, 80% of the sections where users use intelligent driving function are high-speed driving scenarios, which are users’ rigid demands at present. Meanwhile, the definition of high-level products is still not defined clearly yet. The wave of stacking in the industry is unsustainable due to the failure in balancing cost and user experience. Attention in the whole industry shall be focused on high-speed L2+ intelligent driving products, which could satisfy users’ rigid demands with optimal software algorithm performance and extreme safety, and more powerful functions than L2. MAXIPILOT®2.0 is leading in technology performance, and supports lightweight high-speed NOA solution with heavy perception and light map, intelligent functions such as memory line parking, GOD (general object detection), urban area intersection traffic assistance enhancement, etc., bringing users special function enhancement experience. With a focus on users’ rigid demand and eliminating “unnecessary designs”, the product embodies the reasonable matching between product price and user experience, thus improving users' recognition of the intelligent driving product. Meanwhile, the product supports data intelligence system and MET-TOOL simulation development tool technology base, realizing data-driven double closed loop of internal self-iteration and external promotion and upgrading. In the future, efforts will be made to solve the present pain points in the industry, such as high price and poor experience of NOA and driving-parking integration, and to produce highly integrated products with cost performance, so as to satisfy customers’ and users’ higher demands in medium and long-term high-level intelligent driving market.