Easy Car News On January 11, Pony Intelligence officially announced its passenger car intelligent driving business product lines - intelligent driving software brand "Pony Sense Way", domain controller "Fangzai", and data closed-loop tool chain "Cangdao", and has set up an independent business unit (Personnel). "The company has established an independent business unit (Personally Owned Vehicles, or POV) to run the business.
PonySmart: The ultimate experience, cost-effective, nationwide landing smart driving solution
PonySmart solution is based on a number of technological breakthroughs in the field of self-driving algorithms accumulated by PonySmart for many years: PonySmart's own BEV (Bird's Eye View) sensing algorithm, and the POV (Personal Owned Vehicles) tool chain. 's Eye View) perception algorithm, a large model to identify various types of obstacles, lane lines and passable areas and other information, minimizing the need for arithmetic, and at the same time, in the absence of high-precision maps, only using navigation maps can also achieve high-speed and urban NOA function - this is the pony truly This is one of the key capabilities of Pony to truly "know the way". In high-speed scenarios, PonySmart can make extreme use of sensors, using fisheye cameras to participate in the driving BEV perception model, reducing the dependence on the number of sensors - using as low as six cameras (four fisheye cameras and one long-distance camera each in the front and rear directions) and one forward-facing millimeter-wave radar, to create an extremely cost-effective driving and parking integrated solution, the Realize high-speed NOA, memory parking, active safety and other functions, and can be adapted to a variety of chips.
Based on years of accumulation of complex scene processing, PonySmart is the first to create a game interactive planning algorithm, which takes into account the game between self and social vehicles, so that the L4 vehicles have already realized fully unmanned self-driving scale operation in busy urban areas of Beijing and Guangzhou. This set of algorithms gives Pony Sense Way intelligent driving solutions a smooth driving experience like that of an "old driver" - high-speed NOA can realize flexible shuttling on busy urban elevated roads and on-ramps with a greatly streamlined number of sensors, and the human takeover indexes are several times better than those of the mainstream products in the industry; urban NOA can realize flexible shuttling on busy urban elevated roads and on-ramps, and the human takeover indexes are several times better than those of In addition, the urban NOA experience is close to that of L4 Robotaxi, and with a more cost-effective solution, Pony Sense and Way realizes a driving experience that is safe, comfortable, and efficient.
In the traditional development model, several functions are often developed independently and then integrated. PonySmart has subversively pioneered "full-scene ACC", "full-scene LCC", and "full-scene NOA", regardless of the presence or absence of high-precision maps, and regardless of whether you are driving in the urban area, on the highway, or parking, the responsibilities of the different functions are the same as those of the other functions. With or without high-precision maps, whether in urban areas, highway driving or parking, the same set of game interactive planning algorithms is used for the same part of duties in different functions, and the experience is consistent. For example, the longitudinal control and social vehicle cut-in prediction algorithms used in the ACC function are the same as those of LCC and NOA, which can smoothly deal with congested vehicles and realize "zero-pedal" driving; the lateral control algorithms of LCC are the same as those of NOA, which can avoid next-door large vehicles, bypass construction and illegal parking vehicles; and the lane change of the paddle lever of LCC uses the same set of lane change algorithms as those of the autonomous lane change of NOA. LCC uses the same set of lane changing algorithms as NOA's autonomous lane changing algorithms, and has excellent lane changing ability in dense traffic; the parking scenario and urban driving scenario both use the same set of narrow road meeting algorithms to improve the efficiency of traffic, and both can safely and comfortably avoid crossing pedestrians.
In addition, based on its self-developed high-precision positioning algorithm, PonySmart can cooperate with car companies and hardware suppliers to customize the positioning sensors, which further reduces the cost of the entire vehicle hardware under the premise of ensuring performance and experience. PonySmart fully optimizes all modules of the software algorithm to bring better parking performance with the same hardware solution, and breaks the pattern of stacked hardware selection according to the needs of car companies for functions and performance.
At present, based on different sensors and arithmetic configurations, Pony Sense and Way has launched three system-level solutions: PonyClassic, PonyPro and PonyUltra, which provide differentiated reference hardware configurations, and can also be customized according to the customer needs of the overall program.
FangZai: cost-effective, small size, the most suitable for automated driving car-grade domain controller
In addition to the driving experience solutions, PonySmart also provides high-quality, cost-effective intelligent driving hardware module - the The self-developed domain controller "Fangzai" has already started mass production and delivery, and the target customers include automobile enterprises. PonySmart is also actively exploring the application of Fangzai in the field of low-speed unmanned driving, to meet the needs of customers for computing platforms, solve the pain point of high cost of the core hardware, and empower unmanned delivery, sanitation, ports and mining and other application scenarios.
At present, PonySmart has launched the industry's first batch of domain controllers with different configurations of DRIVE Orin chips, including single OrinX chips and dual OrinX chips, and after more than a year's road tests in many places in China, the hardware performance of the square-loaded series of products has been sharpened to the optimal state.
In addition to OrinX, PonySmart is actively laying out and adapting other chips and domain controller platforms, and is cooperating with a number of industrial chain enterprises to enable the large-scale deployment of autonomous vehicles in the future.
PonySmart has accumulated nearly 20 million kilometers of open-road autonomous driving data, experienced multiple rounds of high-performance computing platform iterations and hardware and software system optimization, fully understands the needs of autonomous driving, and is well versed in the requirements of autonomous driving controller design and software adaptation, and has poured this experience into the development of domain controllers. Fangzhai is more suitable for intelligent driving application scenarios in terms of performance, safety, cost, reliability, etc., and better matches the upper layer software algorithms, which greatly improves the overall operational efficiency. At present, Pony Smart domain controller has passed the automotive functional safety ASIL-D development process certification and international quality system certification, all-round to meet the standards of vehicle-grade domain control.
The Dome: A toolchain for solving the data explosion problem and accurately evaluating smart driving systems
Relying on the mature toolchain system polished during the internal development process, PonySmart has launched a data closed-loop toolchain product, "The Dome. The Cangdao data closed-loop tool chain consists of two core modules, namely, the big data platform for vehicle-cloud collaboration and the cloud-based large-scale simulation platform, which, together with the flexible access to data annotation tools and model training tools, can realize full coverage of two types of customers' core needs - the full-volume data closed-loop during the R&D and testing phases, and the data closed-loop during the mass production phase based on the shadow model. The data closure based on shadow model in mass production stage.
PonySmart has supported the Cangdao toolchain to provide software products and services in a modular and pluggable way, and has served a number of automotive companies to support the rapid iteration of smart driving software.
The vehicle-cloud collaborative big data platform in Cangdao's products can efficiently solve the problem of data explosion that is commonly faced by the research and development of current smart driving systems. Through the synergistic mechanism of vehicle and cloud, the platform can accurately mine high-value data, provide data analysis suite, realize one-stop data utilization function, and help customers efficiently complete data collection, storage and mining under limited cost based on y optimized data management system.
The cloud-based large-scale simulation platform in the Cangdao product is the engine for the rapid iteration of the Smart Driving System, which supports fast and accurate multi-dimensional indicator evaluation, covering various aspects such as safety, compliance, comfort, and passage efficiency. The platform is based on a perfectly consistent simulation engine and Smart Agent technology, which maximizes the credibility of the simulation results; with the ability of large-scale task scheduling in the cloud, Cangdao's simulation test can travel 10 million kilometers per day, realizing the effect of software changes in a low-cost way, and guaranteeing the system's ability to be continuously upgraded.
The Cangdao toolchain is the productized manifestation of a complete set of data-driven automatic driving system R&D model and methodology accumulated by PonySmart over the past six years, and it is also the "backbone" of PonySmart's achievement of many technological milestones. The advanced and superiority of this product will help more intelligent driving research and development teams, and empower the automotive industry to move forward with the tide of intelligentization.
EasyCar's Super Evaluation System is now online! The newest addition to the list is the new "Car Review System", which provides professional, hardcore, and comprehensive automotive review content!