Official Amap technology account showcasing all of Amap's technical innovations.
Gaode solved the WebGL‑context limit in AMap by rendering all visual layers on a single canvas with a shared context, disabling autoClear, preserving depth ordering, and using per‑layer EffectComposers merged via a final composer, enabling multiple scenes and post‑processing without exceeding browser limits.
Rate limiting protects system stability by capping request rates, and this article explains five Java algorithms—Fixed Window, Sliding Window, Leaky Bucket, Token Bucket, and Guava's RateLimiter—showing their principles, pros and cons, and providing sample implementations and a Spring @Limit annotation for practical enforcement.
At the fifth 2023 Mapping Technology Forum, Amap’s General Manager Xiang Zhe presented his report on creating a high‑quality ‘Live’ map, describing how the company’s HQ Live Map combines advanced perception with continuously refreshed map data to support intelligent driving and ADAS applications.
By fusing LiDAR point clouds and camera images into a unified bird‑eye‑view space and applying Transformer‑based perception, multi‑sensor fusion, and graph‑diffusion modeling, the proposed BEV+Transformer framework automatically detects and smooths ground‑level line features and signs for high‑definition maps with centimeter‑level accuracy, boosting production efficiency and reducing cost.
Quality assurance is fundamentally about exhaustive falsification to lower online risk, while the testing process seeks optimal ROI by balancing defect discovery against resource investment; a tester’s value hinges on business risk, system maturity, and the ability to become increasingly irreplaceable through deep domain knowledge, proactive risk mitigation, and innovative, end‑to‑end quality solutions.
Vision‑based Bird’s‑Eye‑View (BEV) transforms camera data into a scale‑invariant, geometry‑friendly top‑down map using perspective‑transformation modules such as Lift‑Splat‑Shoot and Pseudo‑LiDAR, incorporates deformable convolutions and attention, and underpins modern autonomous‑driving detectors like BEVDet, BEVDepth, Detr3D, BEVFormer and PETR, while future research targets depth‑estimation bottlenecks, multimodal transformer fusion, and foundation‑model generalization.
The article details Gaode’s journey of building a high‑performance Go ecosystem that scaled from zero to tens of millions of QPS, comparing Go with Java and Erlang, outlining cloud‑native serverless architecture, and sharing real‑world refactoring and optimization case studies such as a million‑QPS rendering gateway and a Go‑based sharding middleware.
The crash in asl::LooperObserverMan::notifyIdle was traced to a compiler‑level store‑store reordering that exposed an uninitialized observer pointer, causing an illegal memory access, and was fixed by inserting a memory barrier or using proper atomic ordering to enforce correct initialization order.
The article broadens software architecture beyond structural design to encompass people, processes, and business value, urging architects to balance realistic, flexible goals, limited resources, commercial‑focused behaviors, and industry trends so they can deliver cost‑effective, scalable solutions that align technology with organizational objectives.
Amap’s lane‑level navigation simulation combines a road‑scene strategy engine—using modified ant‑colony optimization to efficiently cover national and city road networks—with a vehicle‑behavior engine that plans lane‑change paths via Dijkstra and models realistic speed profiles through reaction and curvature intervals, enabling custom high‑precision testing while cutting road‑test costs.
