Microsoft Azure in Autonomous Driving

Transcription

1 Microsoft Azure in Autonomous Driving

2 Microsoft s approach to automotive We complement OEMs and suppliers not compete We ensure your data is always under your control We guarantee that the brand and customer experience belongs to you

3

4 OEM System Engineering Process Engineering Development Product Validation Manufacturing & Service Regional Architecture (s) Feasibility Study/concept Exploration Operations and Maintenance Changes and Upgrades Retirement/ replacement Lifecycle Processes Concepts of Operations System Validation Plan System Validation System Requirements High-Level Design Detailed Design System Verification Plan (System Acceptance) Subsystem Verification Plan (Subsystem Acceptance) Unit/Device Test Plan Unit/Device Testing System Verification & Deployment Subsystem Verification Document/Approval Software/Hardware Development Field Installation Implementation Development Processes Time Line

5 Process, Sample, Reduce Sensor/ Algorithm Testing (Open Loop) Train Control Logic Validation Software in the loop Test Vehicle Ingest/ Store Integrate Build Generate Code F(x) Tag Replay Simulate Performance Simulation Hardware in the loop Render/Convert Test-Drive DATA INGEST & CURATE TEST TRAIN SIMULATE BUILD VALIDATE

6 Cloud and Analytics partner for ACM, an autonomous and smart mobility test facility Leveraged by all OEMs, tier ones and technology start ups Engaged with ACM to influence standards Open source AD Platform 200+ Member Consortium Microsoft is the cloud provider for Project Apollo worldwide with the exception of China Industry OpenADx Interoperable Eclipse Framework Academia Government

7

8 LG s autonomous vehicle program had unique requirements, including portability, security, and fast turnaround time; Data Box Disk was the perfect solution. We needed a way to transfer massive amounts of data for our autonomous vehicle projects, based all around the world. The solution needed to be portable, simple to use, cost-effective and, of course, very secure. The Azure Data Box Disk met all of those criteria. Hyoyuel (Andy) Kim, Senior Manager, LG Electronics. Products and services Azure Data Box Disk Azure Blob Storage Organization size 91,000 Industry Automotive electronics Country Worldwide

9 Process, Sample, Reduce Sensor/ Algorithm Testing (Open Loop) Train Control Logic Validation Software in the loop Test Vehicle Ingest/ Store Integrate Build Generate Code F(x) Tag Replay Simulate Performance Simulation Hardware in the loop Render/Convert Test-Drive DATA INGEST & CURATE TEST TRAIN SIMULATE BUILD VALIDATE Azure Express Route Azure Compute Databricks & HDInsight Azure IoT Edge Azure Storage, Data Lake Gen2 Azure Data Box Azure Batch AI Cycle Cloud Azure Machine Learning Cognitive Services Container and Kubernetes, Visual Studio Team Services Azure Key Vault, AAD

10 SOFTWARE TOOLS AND DEVOPS PLATFORM Machine Learning Source Control CI/CD flows Container Registry Azure Container Service DATA INGEST Data Factory Machine Learning Virtual machine In-car data storage Async upload Azure blob storage (cool, archive) ML HDInsight, Azure Databricks CosmosDB Logic App Kubernetes Service Azure GPU VM Simplygon Cycle Cloud PowerBI Cosmos DB Data reduction, redaction Fast upload Stream Analytics Advanced Analytics HDInsight Databricks Batch Service Bus IoT Edge ADLS Gen 2 and Azure blob storage Hot Tier

11 Process, Sample, Reduce Sensor/ Algorithm Testing (Open Loop) Train Control Logic Validation Software in the loop Test Vehicle Ingest/ Store Integrate Build Generate Code F(x) Tag Replay Simulate Performance Simulation Hardware in the loop Render/Convert Test-Drive DATA INGEST & CURATE TEST TRAIN SIMULATE BUILD VALIDATE

12 Offline Data Transfer Online Data Transfer Data Box Data Box Disk Data Box Heavy Data Box Gateway Data Box Edge Capacity: 100 TB Weight: ~50 lbs Secure, ruggedized appliance Data Box enables bulk migration to Azure when network isn t an option. Capacity: 8TB ea.; 40TB/order Secure, ruggedized USB drives orderable in packs of 5 (up to 40TB). Currently in Preview Perfect for projects that require a smaller form factor, e.g., autonomous vehicles. Capacity: 1000TB Weight 500+ lbs Secure, ruggedized appliance Currently in Preview Same service as Data Box, but targeted to petabytesized datasets. Virtual device provisioned in your hypervisor Supports storage gateway, SMB, NFS, Azure blob, files Currently in preview Virtual network transfer appliance (VM), runs on your choice of hardware. Local Cache Capacity: ~12 TB Includes Data Box Gateway and Azure IoT Edge. Currently in preview Data Box Edge manages uploads to Azure and can pre-process data prior to upload. Network Data Transfer Edge Compute Order Send Fill Return Upload Cloud to Edge Edge to Cloud Pre-processing ML Inferencing

13 Massive Data Ingest ExpressRoute Ingest up to PB of data daily with Azure Networking Express Route Service Private, high speed network connections Connecting your data center and Azure Predictable performance and Secure Express Route partners

14

15 Data Storage Azure Blob Storage Store up to Exabytes of data in massively scalable object storage for unstructured data Scalable Foundational service for Microsoft >40 million transactions per second Multi-PB accounts Frequently accessed data Lowest transaction cost Immediate (ms) Less frequently accessed data Lower capacity cost Immediate (ms) Rarely accessed data Lowest capacity cost Hours Performant Secure & Compliant Durable 50+Gbps account throughput Continued enhancements under development Client & Service Encryption AAD Integration + ACLs Broad & deep compliance portfolio Multiple redundancy options Strong consistency, data integrity Policy: Versioning & WORM locks Cost Effective Integrated storage tiers Lifecycle management Rich metrics

16 Identity & access management Data protection Network security Threat protection Security management Azure Active Director y Encr yption (Disks, Storage, SQL) VNET, VPN, NSG Azure Security Center Multi-Factor Authentication Azure Key Vault Application Gateway (WAF), Azure Firewall Microsof t Antimalware for Azure Azure Log Analytics Role Based Access Control DDoS Protection Standard Azure Active Director y (Identity Protection) ExpressRoute + Partner Solutions

17 Partner-based Solutions Sensor data fusion tools Data Extraction Data Preparation Microsoft Services Annotation Data extraction Labeling and annotation Data analysis Blob Cognitive Services Services Logic Apps Functions HDInsight Search Databricks Cosmos DB Batch Data selection and transformation, PII data redaction, annotation and training data preparation Data anonymization Workflow orchestration and managed services

18 Analyze your Data in one place Big Data Use Cases Ingest & ETL Streaming Analytics & Machine Learning Data Aggregation Presentation Monitor App Insights Log Analytics Data Factory Event Hubs IoT Hub Stream Analytics Functions Batch Data Lake Analytics Machine Learning Data Warehouse Search CDN Power BI Azure HDInsight Blob Storage Pillars Open & Interoperable Manageable & Cost Efficient Scalable & Performant Secure & Compliant Durable & Available Manageable & Cost Efficient Scalable & Performant Secure & Compliant

19 Deploy AI models in Azure for autolabeling images to optimize for speed of labeling and possibly cost Semantic Segmentation Each Pixel of the image is assigned a category Object Detection & Classification Bounding box drawn around each object of interest 3D Point Cloud Labeling Objects of interest as assigned a category in 3D LIDAR point cloud

20 Process, Sample, Reduce Sensor/ Algorithm Testing (Open Loop) Train Control Logic Validation Software in the loop Test Vehicle Ingest/ Store Integrate Build Generate Code F(x) Tag Replay Simulate Performance Simulation Hardware in the loop Render/Convert Test-Drive DATA INGEST & CURATE TEST TRAIN SIMULATE BUILD VALIDATE

21 Partnership Opportunities Open loop testing (e.g. ADTF) Test repository Scene Reconstruction and Replay Algorithm scoring Results Analysis Simulation tools for sensor and algorithm validation Microsoft Services Workflow Management Services Cosmos Blob Cycle Cloud CPU DB Compute Logic Apps Batch Key Vault Service Bus PowerBI Comprehensive Validation framework Verification and validation of training algorithms and sensors via open loop testing

22

23

24 Build and deploy deep learning models Caffe2 Azure databricks Keras MS cognitive toolkit Azure databricks Scale out clusters TensorFlow Notebooks Azure ML Services Batch AI Scale out clusters Streamline AI development efforts Leverage popular deep learning toolkits Develop your language of choice Scale compute resources in any environment Choose VMs for your modeling needs Process video using GPU-based VMs Leverage your favorite deep learning frameworks Quickly evaluate and identify the right model Run experiments in parallel Provision resources automatically TensorFlow MS Cognitive Toolkit PyTorch Scikit-Learn ONNX Caffe2 MXNet Chainer

25 Operationalize and manage models with ease Azure databricks Azure ML services AKS ACI Containers Train and evaluate models Model MGMT, experimentation, and run history IoT edge Bring models to life quickly Build and deploy models in minutes Iterate quickly on serverless infrastructure Easily change environments Proactively manage model performance Identify and promote your best models Capture model telemetry Retrain models with APIs Deploy models closer to your data Deploy models anywhere Scale out to containers Infuse intelligence into the IoT edge

26 Simulation is the fastest way to drive the billions of miles needed to discover the breadth of real-world anomalies required to develop AD solutions. Sensor Simulation Scenario Simulation Engineering Simulation Photo-realistic rendering of modeled sensors (Lidar, radar, RGB cameras) for the training and validation of Perception AI algorithms. Render once, test many times Scenario definition and simulation for path planning and navigational maneuvers (overtake, parallel parking, etc) using post perception object level simulation. Run 7200 simulated scenarios covering ~5000 miles per day/per node Mathematical based simulation for Structural CAD design, Fluid dynamics, Thermo dynamics, material properties Design sensor properties and placement on vehicle options at scale before hardware prototyping is needed. Simulate accidents and AD navigation to a safe stop

27 Simulation/Reinforcement Learning Centric Model Render/convert Simulation Engine Test Vehicle Fleet Ingest/Store Sensor Fusion, Process, Sample, Reduce, Tag, QA Training Dataset Train Algorithm Validation Engine Build Integrate Control Logic Validation Generate Code Hardware in the loop Production Vehicle Fleet Performance Simulation Software in the loop F(x) Test-Drive Drive Millions, Train on Billions

28 Process, Sample, Reduce Sensor/ Algorithm Testing (Open Loop) Train Control Logic Validation Software in the loop Test Vehicle Ingest/ Store Integrate Build Generate Code F(x) Tag Replay Simulate Performance Simulation Hardware in the loop Render/Convert Test-Drive DATA INGEST & CURATE TEST TRAIN SIMULATE BUILD VALIDATE

29 Control Logic Validation Software in the loop Partner-based Solutions Performance Simulation Generate Code F(x) Hardware in the loop Comprehensive test management framework HIL solutions Test-Drive System validation tools Microsoft Services Workflow management services Managed services Blob storage Batch GPU VM Active Directory Container service Embedded system validation via hardware-inloop and software-in-loop

30 Copyright Microsoft Corporation. All rights reserved.