The LEADER framework achieves ten‑millisecond "eye‑open" LiDAR relocalization while surpassing the decimeter‑level accuracy of classic retrieval‑registration pipelines, using cylindrical projection, sparse convolution, and a Truncated Relative Reliability loss, as demonstrated on the NCLT benchmark.
This comprehensive review surveys how efficient modeling paradigms, architecture designs, and inference algorithms can overcome the compute‑speed trade‑off in video world models, and examines their impact on autonomous driving, embodied AI, and interactive game simulations.
World Engine introduces a post‑training pipeline that combines high‑fidelity 3DGS simulation, hard‑case mining with diffusion generation, and reinforcement‑learning optimization to give autonomous‑driving models true decision‑making ability, surpassing data‑scaling limits and achieving significant safety gains in both industrial simulations and real‑world tests.
ColaVLA replaces explicit text‑based reasoning with latent‑space inference and a hierarchical parallel planner, achieving lower trajectory error, reduced collision rates and up to ten‑fold faster inference while preserving safety and real‑time performance in autonomous driving benchmarks.
This article details how a leading autonomous‑driving company tackled multi‑tenant resource contention, low GPU utilization, and distributed task dead‑locks on a heterogeneous Kubernetes cluster by adopting Koordinator’s ElasticQuota, Reservation, Gang and Device‑Share features, achieving higher allocation rates, better fairness, and significantly improved GPU efficiency.
The DiffVL framework redefines visual localization as a diffusion‑based GPS denoising task, using BEV‑conditioned visual cues and standard SD maps to achieve sub‑meter accuracy without high‑definition maps, and demonstrates its superiority through extensive autonomous‑driving experiments.
UniMapGen introduces a generative, multimodal framework that models lane lines as token sequences, employs an iterative state‑update mechanism for global consistency, and achieves state‑of‑the‑art performance on large‑scale satellite‑derived map construction, enabling seamless lane‑level navigation worldwide.
This article walks through a complete data‑science pipeline for autonomous driving, covering CSV/JSON preprocessing with pandas, rule‑based scene classification, a weighted complexity‑risk model, and visual analytics using Matplotlib and Seaborn to identify high‑risk driving scenarios.
At CES 2026, Jensen Huang presented NVIDIA's Rubin platform, highlighting five hardware innovations that cut inference token cost tenfold and reduce GPU requirements fourfold, while also launching a suite of open‑source models for Agentic AI, robotics, autonomous driving and AI‑for‑Science, drawing praise from industry leaders.
This roundup presents five recent AI papers covering hierarchical sparse attention for ultra‑long context, Nvidia's Alpamayo‑R1 VLA model for autonomous driving, the non‑autoregressive F5‑TTS system, LatentMAS for latent‑space multi‑agent collaboration, and Deeper‑GXX that deepens arbitrary graph neural networks, highlighting each method's key innovations and reported performance gains.
Baidu Apollo has donated L4‑level autonomous vehicles to Fudan and Tongji universities, supporting AI research, talent cultivation, and industry‑academia integration through the OnSite competition and comprehensive educational programs, while highlighting the broader impact of autonomous driving on the AI ecosystem in China.
This comprehensive analysis examines the strategic importance, technical challenges, innovation pathways, and market landscape of domestic AI chips in China, highlighting key players, regional clusters, core applications such as intelligent computing, autonomous driving, and robotics, and projecting future industry bottlenecks and opportunities.
Gaode’s lane‑level map technology evolves from LiDAR‑based methods to multimodal large‑model generation, delivering meter‑level navigation, empowering advanced driver‑assistance and autonomous driving, while its RoadGPT model and recent research breakthroughs illustrate a new era of dynamic, intelligent spatial intelligence.
FSDrive introduces a spatio‑temporal chain‑of‑thought approach that enables visual language models to generate future driving scenes as images, improving trajectory planning accuracy and safety by eliminating cross‑modal gaps and enforcing physical constraints in autonomous driving.
This article shares practical experience training perception‑detection mini‑models on two different compute cards, covering environment setup, technical architecture, common dependency issues, performance‑boosting tricks such as CPU process pools, torch dataloader tuning, NCCL P2P handling, and CPFS storage optimization.
This tutorial walks through installing gym and highway‑env, configuring six driving scenarios, processing observations (kinematics, images, occupancy grids), defining actions and rewards, constructing a DQN network, training it with a reinforcement‑learning loop, and analyzing collision, time, and reward metrics.
This article reviews the V2X‑ViT collaborative perception framework for autonomous driving, detailing its end‑to‑end pipeline, the novel HMSA and MSwin attention mechanisms, and the delay‑aware positional encoding that together enable high‑accuracy 3D object detection across vehicles and infrastructure.
SeqGrowGraph introduces a novel chain-of-graph expansion framework that incrementally builds lane topology graphs using a Transformer-based autoregressive model, achieving state‑of‑the‑art performance on large autonomous‑driving datasets such as nuScenes and Argoverse 2 by accurately modeling complex road structures.
The article argues that despite rapid advances in large‑model benchmarks, the lack of real‑world data and suitable tasks creates a fundamental gap that will keep AI agents far from replacing 80% of white‑collar work for many years, making hype about imminent AGI unrealistic.
This article explores Zheng Naning's lecture on machine behavior, embodied intelligence, and their challenges, outlining AI development stages, the need for explainable and cooperative machine actions, and the specific hurdles and frameworks for achieving safe, adaptive autonomous driving in dynamic environments.
IDC’s April 2024 report shows China’s automotive cloud market reaching 6.51 billion RMB in the second half of 2024, with a 27.4% YoY growth, and highlights Baidu Cloud’s 34.5% share in the 16.34 billion RMB autonomous‑driving solution market, driven by AI advances and expanding compute investments.
A March 2025 crash involving a Xiaomi‑branded autonomous vehicle illustrates how a four‑second algorithmic decision loop, inadequate night‑vision sensors, flawed handover timing, and poor emergency‑exit design combined to create a lethal scenario that exposes the deadly risks of over‑relying on L2 driver‑assist systems.
This article explores how high cardinality in Prometheus metrics impacts AI large‑model and autonomous‑driving observability, explains the underlying concepts, outlines the performance and cost challenges, and presents practical design, collection, and query‑side solutions—including metric modeling, pre‑aggregation, and remote‑read pushdown—to keep monitoring efficient and scalable.
This article shares practical experience from Volcengine's managed Prometheus service and its deep integration with large‑model and autonomous‑driving platforms, explaining what high cardinality is, its impact on monitoring systems, root causes, and a range of design, collection, and analysis techniques to mitigate it.
Gaode Map and Xi'an Jiaotong University introduce the “Driving by the Rules” task, releasing the MapDR benchmark that integrates lane‑level traffic‑sign regulations into online‑constructed HD maps, and provide modular (VLE‑MEE) and end‑to‑end (RuleVLM) baselines to evaluate rule extraction and lane association.
Traditional IT architectures suffer from manual, passive operations and limited scalability, prompting a shift toward adaptive intelligent architectures that leverage neural architecture search, elastic networks, and meta‑learning to dynamically adjust models across domains such as autonomous driving, mobile devices, robotics, and personalized recommendation, while addressing efficiency, generalization, and real‑time challenges.
This tutorial explains how to install the gym and highway‑env packages, configure a highway simulation environment, process observations and actions, build a DQN network in PyTorch, train the agent, and analyze training results for autonomous driving scenarios.
From the 1925 driverless “American Wonder” to today’s AI‑driven robotaxi wars, the article traces the historic roots, recent breakthroughs by Waymo, Tesla and Baidu, and analyzes China’s Luo Bo platform, market forecasts, competitive dynamics, and the strategic challenges facing the global autonomous‑driving industry.
The article discusses Baidu’s new autonomous‑driving service “Robo Fast Run,” its potential impact on ride‑hailing jobs, the technical challenges involved, and provides detailed guidance on the qualifications and learning paths for engineers seeking positions in Baidu’s intelligent‑driving divisions.
This article details how Alluxio was adopted to replace NAS in autonomous driving model training, describing the data closed‑loop workflow, the challenges of the previous system, Alluxio's architectural benefits, deployment strategies across single and multiple data centers, functional and performance testing, operational tuning, and the resulting cost and efficiency gains.
This article explains how to install the gym and highway‑env packages, configure the environment for various driving scenarios, define observations, actions and rewards, build a DQN network in PyTorch, run the training loop, and analyze the resulting performance metrics.
The presentation by NIO senior technology planning expert Chen Jiong explores the development trends of intelligent electric vehicles, showcases how large AI models empower various automotive scenarios, and shares NIO's practical implementations, offering insights on industry-focused solutions, problem‑driven application, and unified architecture design.
Baidu unveiled Apollo Open Platform 9.0 on December 19, delivering Package Management 2.0, new perception models with 4D millimeter‑wave radar support, incremental training, and a revamped Dreamview+ UI, and will host two live‑stream courses on December 27‑28 to explain the PnC and perception upgrades, with gifts for registrants.
This article analyzes Baidu’s comprehensive autonomous‑driving toolchain, covering its evolution from research to mass‑production, real‑world case studies, compliance and data challenges, and the integrated cloud, AI, and simulation services that enable OEMs to accelerate smart‑vehicle deployment.
This article summarizes Baidu’s senior manager Xu Peng’s presentation on the evolution from R&D‑focused to production‑ready autonomous driving toolchains, highlighting cloud simulation, data‑closed‑loop, AI‑driven labeling, compliance, efficiency, service, and cost challenges, and outlining Baidu’s integrated solutions for the automotive industry.
On September 22, the self‑designed "Qi Ji" A2 autonomous vehicle for smart‑city data production was inspected by China Merchants Group Vehicle Technology Research Institute, passed all 11 major test items of the T/CSAE 285‑2022 functional‑type unmanned vehicle field test standard, and the detailed results are presented in this report.
At the 2023 Baidu Cloud Intelligence Conference, Baidu AI Cloud outlined a comprehensive, four‑layer solution—spanning distributed cloud infrastructure, AI‑focused compute, data compliance, and end‑to‑end toolchains—to address the challenges of electric, intelligent vehicles, large‑model deployment, and regulatory compliance in autonomous driving.
This article reviews the background and challenges of data security in wireless communications, introduces Trustworthy Blockchain-based Federated Learning (TBFL), details a two‑layer TBFL architecture with edge computing, discusses its features, key technologies, and autonomous‑driving applications, and outlines current limitations and future research directions.
Elon Musk streamed a 45‑minute, unedited live demonstration of Tesla’s new Full Self‑Driving V12 system, revealing its end‑to‑end AI video‑training approach, hardware constraints, power efficiency, code reduction, and real‑world performance across varied road scenarios, while highlighting a brief manual intervention.
This article explains how to install gym and highway‑env, configure the environment, process state, action and reward data, build a DQN model in PyTorch, run training loops, and analyze results for autonomous driving simulations using reinforcement learning.
At a June 2023 forum in Beijing, Didi partnered with leading Chinese universities to showcase award‑winning projects that applied deep‑learning positioning, long‑term pricing models, high‑precision two‑wheel navigation, in‑vehicle voice separation, and data‑driven autonomous‑driving, illustrating how academia‑industry collaboration turns cutting‑edge research into commercial value.
In 2022 Baidu’s open‑source ecosystem surged past 1,000 GitHub projects and 20,000 contributors, with flagship initiatives like PaddlePaddle, Apollo 8.0, XuperCore, Apache Doris, bRPC, Teaclave, BFE, Baetyl, HugeGraph and amis achieving major releases, graduations, and industry recognitions.
Baidu’s Baige AI heterogeneous computing platform delivers an end‑to‑end, low‑cost AI IaaS for autonomous driving, covering data cloud, tiered storage, RapidFS caching, AIAK‑Inference and AIAK‑Training acceleration, GPU container virtualization, and remote GPU pooling, achieving up to 5× faster data access, 391% training speedup, 90% inference latency reduction, and 60% simulation cost cut.
This article outlines Baidu Intelligent Cloud’s end‑to‑end autonomous driving platform, detailing its AI foundation, massive cloud‑based data and compute requirements, flexible deployment strategies for various manufacturers, and comprehensive toolchains for data collection, annotation, training, simulation, and compliance.
Baidu’s cloud‑based simulation platform now covers the entire autonomous‑driving development cycle, offering integrated testing for perception, decision, planning and control, dramatically reducing iteration time, boosting scenario coverage to 98% and achieving a 99.8% model deployment success rate.
This article outlines Baidu Intelligent Cloud's comprehensive, low‑cost solution for autonomous‑driving data pipelines—from road data collection and compliance, through annotation, management, and model training, to simulation—highlighting the platform's tools, services, and security measures that accelerate development.
The talk outlines Baidu’s Baige AI IaaS solution for autonomous driving, detailing a low‑cost, high‑efficiency cloud stack that accelerates data access fivefold, boosts model training speed up to 391 %, cuts inference latency by 90 %, reduces simulation costs by 60 %, and explains the underlying storage, compute, container and GPU virtualization technologies.
This article presents a comprehensive overview of several publicly available autonomous driving datasets, focusing on Series Six of the Integer Intelligence collection, which includes StreetLearn, UTBM RoboCar, Multi‑Vehicle Stereo Event Camera, comma2k19, the Annotated Laser Dataset, Ford, and Oxford RobotCar, detailing their sources, download links, publication years, key features, and research relevance.
The article reviews the TVP 5G technology sharing conference, summarizing expert analyses on 5G’s three‑year commercial rollout, its impact on autonomous driving, industrial IoT, mining and port logistics, the challenges of large‑scale deployment, and strategic recommendations for accelerating 5G‑driven digital transformation.
This tutorial explains how to install the gym and highway‑env packages, configure the simulation environment, process state and action representations, implement a DQN network in PyTorch, and train the model while visualizing performance metrics for autonomous driving tasks.
Cloud computing delivers on‑demand, pay‑as‑you‑go resources through IaaS, PaaS, and SaaS layers, while edge computing extends these services toward devices to meet latency, bandwidth, and privacy demands, using lightweight K3s‑based clusters that address management, connectivity, and resource constraints for applications such as autonomous driving.
In a newly released audio course, Baidu founder Li Yanhong outlines how China could become the first country to commercialize large‑scale L4‑L5 autonomous driving, emphasizing vehicle‑road collaboration, AI‑enabled traffic control, and the broader impact on intelligent cities and society.
The article examines the strikingly similar development trajectories of electric vehicles and distributed databases, tracing their historical origins, technological challenges, market dynamics, and recent breakthroughs such as Tesla's autonomous electric cars and OceanBase's native HTAP capabilities, while highlighting the role of open‑source and scaling strategies.
This article provides a comprehensive overview of GPU technology, covering its multi‑core architecture, market oligopoly among Intel, NVIDIA and AMD, classifications of integrated and independent GPUs, and the three major application trends of gaming performance, artificial intelligence/deep learning, and autonomous driving.
Autonomous driving combines real‑time sensor perception, AI‑driven decision making, and electronic control to navigate vehicles, with hardware such as cameras, LiDAR, and chips from firms like Mobileye and NVIDIA, software stacks covering mapping, localization, planning, and safety standards, progressing from Level 1 assistance to full Level 5 autonomy while reshaping transportation through electrification, sharing, connectivity, and supportive policies.
A roundup of recent tech headlines covers Xiaomi's logo stance, a temple investment rumor denial, Elon Musk's admission about FSD, a job scam alert, Zhihu's comment outage, Mercedes' highway autopilot plans, and a Chinese university's quantum machine learning breakthrough.
This article explains the concept, architecture, and key components of a high‑definition map data distribution engine for advanced driver assistance systems, detailing map precision requirements, model abstractions, synchronization protocols, integration options, quality assurance methods, and typical autonomous‑driving applications.
The Amap Technology Open Day video replay features Xiang Zhe presenting the current status and future of high‑precision maps for autonomous driving, covering autonomous‑driving evolution, map architecture, data collection, processing, production and update technologies, while offering contact details, a 20‑minute Q&A, and information on internship and full‑time positions.
The High‑Definition Map Data Distribution Engine (AHP) delivers centimeter‑accurate, machine‑oriented road‑network and attribute data to ADAS and autonomous‑driving modules via the ADASIS electronic horizon, using layered architecture, path/offset models, attribute interpolation, synchronized control messages, and flexible integration, testing, and future‑ready extensions.
This article introduces Baidu EasyDL's zero‑threshold AI training camp, focusing on an automatic‑driving scene image‑segmentation case study, covering platform features, model selection, data collection, cleaning, annotation, smart labeling, augmentation, deployment, and a live Q&A session.
Gaode’s high‑precision map platform, described by GM Xiang Zhe, details a three‑stage production pipeline, multi‑layer map architecture, and tiered data‑collection strategy that together address city‑road challenges, ensure map freshness, advance positioning and perception algorithms, and support commercial Level‑4 autonomous‑driving deployments.
This article examines the fundamental architecture of autonomous driving, highlighting the three core technical contradictions—rapid iteration versus functional safety, sensor and compute demands, and hardware performance versus automotive-grade safety—while outlining a staged development roadmap, hardware and software evolution strategies, and the long‑term goal of safe, reliable driverless operation.
In a comprehensive interview, Elon Musk shares his views on artificial intelligence, the future of autonomous vehicles, battery resource challenges, the strategic importance of Tesla's Berlin factory, his personal wealth philosophy, the need for multi‑planetary humanity, and his definition of himself as an engineer.
On January 9, 2021, Meituan hosted an unmanned‑delivery technical salon in Beijing where experts presented cutting‑edge AI research—including the CenterMask instance‑segmentation method, 3D geometry‑aware camera localization, multi‑agent trajectory prediction with attention‑based spatio‑temporal graphs, real‑time stereo visual‑inertial odometry calibration, and self‑supervised depth‑pose learning for dynamic scenes.
The paper presents an integrated lateral‑longitudinal control framework for autonomous vehicles that employs a coupled vehicle model and joint constraints within a linear time‑varying model predictive control scheme, yielding a unified performance index and demonstrating more human‑like, balanced tracking of speed, position, and yaw compared with traditional separated controllers.
This article analyzes the evolution of high‑precision positioning technologies—from basic GNSS and sensor‑based dead‑reckoning to multi‑sensor fusion, RTK, visual SLAM, and tightly‑coupled SLAM—explaining how they support lane‑level navigation and advanced driver assistance on both mobile and vehicle platforms.
Huawei has established a wholly‑owned electric‑technology subsidiary with a 250 million RMB capital, expanding its automotive R&D, smart vehicle equipment sales, and patents on autonomous driving, while highlighting a decade‑long push into car‑connected AI chips and strategic investments.
The article details Amap’s (Gaode) technical approach to building and deploying high‑precision maps for autonomous driving, covering accuracy requirements, data collection, point‑cloud alignment, AI‑driven perception and map‑update pipelines, and the challenges of scale, cost, and freshness.
This article explains the concepts of the Internet of Things and big data, explores how massive sensor data fuels smart transportation and vehicle networking, outlines practical applications such as real‑time traffic control and autonomous driving, and analyzes the technical and managerial bottlenecks hindering future growth.
This article examines how 5G‑enabled Multi‑Access Edge Computing (MEC) and Tencent’s TMEC platform, built on TKEStack, deliver low‑latency services such as high‑precision positioning, video processing, QoS control and 5G slicing to support autonomous driving, V2X, cloud gaming and other emerging use cases.
Alibaba is advancing cargo‑focused autonomous driving by building a fine‑grained scenario library and the AutoDrive platform to automate perception, planning, and control algorithms, targeting low‑speed L4 logistics robots while navigating sensor, hardware, and regulatory challenges and anticipating broader L2 adoption and tighter hardware‑software co‑design.
Alibaba’s AutoDrive platform tackles autonomous driving challenges by combining fine-grained scenario classification, targeted algorithm development, and cloud‑based automation to accelerate L4 and logistics‑focused self‑driving solutions, while outlining industry standards, sensor choices, and future research directions.
This article reviews Pony.ai's autonomous driving hardware evolution, detailing the company's hardware milestones, team structure, the Pony Alpha2 system, and the key challenges of cost, power consumption, rapid iteration, mass production, and complex road scenarios, while sharing practical solutions and future directions.
In a live Q&A, Alibaba Amap’s chief scientist Ren Xiaofeng explained how computer‑vision algorithms underpin high‑precision map creation, AR navigation, visual localization and sensor fusion, discussed current hardware limits, deep‑learning bottlenecks, 5G’s role, edge‑cloud cooperation, and offered career advice for transitioning researchers.
This article systematically explains the key technologies behind mobile and vehicle navigation positioning, covering sensor fusion, AHRS, map‑matching algorithms based on hidden Markov models, Kalman filtering, and the evolution toward lane‑level and centimeter‑level accuracy for autonomous driving.
This article explains the principles, components, generation process, and challenges of high‑precision topological and point‑cloud maps, and describes satellite‑based, map‑based, and fused high‑precision localization methods that underpin perception, prediction, planning, and control in autonomous driving systems.
The article reviews the complex system model of autonomous heavy trucks, outlines traditional and modern planning and control methods—including rule‑based FSM, POMDP, learning‑based and optimization techniques—highlights safety, efficiency, fuel‑economy, and dynamic modeling challenges specific to heavy‑truck and trailer configurations, and shares practical attempts such as lane‑changing, merging, and trailer‑aware trajectory planning.
This article reviews perception technologies used in autonomous heavy‑truck systems—including lane‑line detection, obstacle detection, and LiDAR sensing—detailing traditional and deep‑learning approaches, practical challenges on high‑speed highways, and the cost, performance, and reliability issues faced when moving these solutions to mass production.
This article reviews the rapid development of deep learning, its pivotal role in autonomous driving, outlines end‑to‑end perception‑to‑control pipelines, discusses the strengths and limitations of deep models, and proposes practical strategies such as task decomposition, multi‑method fusion, and sensor integration to improve safety and interpretability.
This article presents a comprehensive overview of L4 autonomous driving heavy trucks, covering system architecture, sensor and computing hardware, data and model platforms, production challenges, safety considerations, and strategies for achieving reliable, high‑performance mass‑produced autonomous trucks.
Zhang Yaqin, a distinguished AI pioneer and former Baidu and Microsoft executive, has joined Tsinghua University as an Intelligent Science Chair Professor, where he will lead research on autonomous driving, AI‑IoT integration, and the establishment of the university’s Intelligent Industry Research Institute.
This article reviews the evolution of autonomous‑driving hardware, discusses key sensor technologies such as LiDAR and GNSS/IMU, outlines mechanical and electronic challenges—including size, weight, temperature, vibration, and electromagnetic interference—and presents Pony.ai’s PonyAlpha platform as a practical solution.
This article explains how Amap leverages computer vision to collect, process, and enhance map data and to deliver low‑cost, real‑time AR navigation, detailing the technical challenges, algorithmic solutions, and the broader mission of connecting the physical world.
This article outlines Pony.ai's comprehensive autonomous driving infrastructure, describing traditional internet back‑end components, additional vehicle‑mounted systems, large‑scale simulation, data challenges, and the reliability, performance, and flexibility practices needed to support rapid growth and safe robotaxi operations.
The article explains how cloud computing’s centralized model faces latency and bandwidth limits, introduces edge and mobile edge computing as complementary solutions, outlines their technical benefits, and examines their critical role in autonomous driving and upcoming 5G ecosystems.
This article explores how autonomous driving systems can simultaneously ensure safety and passenger comfort by optimizing planning and control modules, defining safety and comfort metrics, formulating constraints and cost functions, and employing models such as the bicycle model for lateral and longitudinal control.
This article traces the development of automotive electronic control units from early CPU‑centric ECUs to centralized domain controllers, examines the rise of GPU‑based AI accelerators for assisted driving, and explains why ASICs are expected to dominate future autonomous‑driving chips, while profiling key industry players and their strategies.
This article explains how a conventional vehicle can be progressively upgraded through hardware retrofits, sensor integration, mapping, perception, control, and planning modules to achieve SAE Level 4/5 autonomy, using a step‑by‑step analogy with driver training and iterative testing.
Pony.ai shares its extensive experience on multi‑sensor perception for autonomous trucks, explaining why sensor fusion is needed, the essential motion‑compensation and calibration steps, and practical camera‑lidar and radar‑lidar fusion techniques that improve detection range and robustness.
The article describes a workflow for generating lane‑level connection relationships in high‑precision navigation data using the NDS standard, detailing how lane groups and connector IDs are assigned uniquely within tiles and across a 9‑tile neighborhood for both NDS 2.5.2 and 2.5.4 versions.
The article reviews the essential infrastructure for autonomous driving, covering massive sensor data storage strategies, the role of metadata, offline and real‑time computing platforms, basic micro‑service components, and various business scenarios, highlighting why robust big‑data handling is critical.
The article traces the evolution of autonomous driving from early concepts to modern implementations, highlighting Pony.ai’s technical innovations in sensor fusion, high‑definition mapping, simulation, data processing, software iteration, and the challenges of scaling vehicle fleets for commercial deployment.
The article explains how visual‑inertial sensor fusion, combined with GNSS and LiDAR, enables large‑scale production of high‑precision maps, detailing hardware choices, processing pipelines, Gaode's implementation, current challenges, and future directions toward multi‑source data integration.
This article outlines the high‑precision positioning requirements for autonomous vehicles and reviews the core technologies—including INS, GNSS/RTK, high‑definition maps, wheel sensors, and multi‑sensor fusion with Kalman filtering—detailing their principles, challenges, and typical deployment scenarios.
This article outlines Pony.ai's perception pipeline, comparing traditional and deep‑learning approaches for 2D and 3D object detection, detailing sensor fusion, detection methods, challenges such as occlusion and distance estimation, and how hybrid techniques improve accuracy for autonomous driving.
At CVPR 2019, Didi’s technology team co‑hosted an autonomous‑driving workshop, showcased the D²‑City dataset, and secured second place in the AI City Challenge by introducing a modular multi‑camera tracking framework, a CNN‑based single‑camera tracker, and a staged aggregation strategy, while outlining its hybrid dispatch commercial plan.
The talk explains how massive, well‑annotated data fuels autonomous‑driving AI, covering data annotation metrics, team structure, efficiency‑boosting techniques, system stability, and how data‑driven development and decision‑making improve model training, evaluation, and product priorities.
The article explains cloud computing fundamentals, highlights its limitations, introduces edge and mobile edge computing (MEC) as complementary solutions, and details how MEC reduces latency and bandwidth usage to enable advanced use cases such as autonomous driving in the 5G era.
The article presents a comprehensive overview of Pony.ai's autonomous driving infrastructure, covering the core infrastructure team’s responsibilities, vehicle onboard systems, simulation platform, data architecture, and supporting services, while discussing the technical challenges and engineering practices employed to achieve scalability, reliability, and high performance.
This article presents a comprehensive overview of high‑definition mapping for autonomous vehicles, covering topological and 3D grid maps, the data‑collection and processing pipeline, key challenges such as cost and scalability, and detailed discussions of SLAM, pose‑graph optimization, ICP, and multi‑sensor localization techniques.
