Autonomous Vehicle Software Image

Autonomous Vehicle Software

Purchase

The development of a robust Autonomous Vehicle software stack is a highly complex engineering task, requiring automakers and their suppliers to develop software that can perceive and comprehend the environment, predict the behavior of dynamic agents within the scene, and execute maneuvers in a way that does not contribute to an unsafe scenario and does not cause the occupant any discomfort.

This report investigates the artificial intelligence and software development techniques being adopted by the automotive industry too meet the challenge, and assess issues such as early vs. late sensor fusion, and how the correct balance of deterministic and trained software can help to address functional safety requirements, and give confidence that no autonomous vehicle will ever be the responsible party in an accident.

As well as considering the technical challenges which remain in autonomous vehicle software development, the report also addresses how this software can be brought to market and monetized, comparing the more flexible and modular approaches of vendors such as AIMotive and NVIDIA with the more integrated, end-to-end approaches of Mobileye and Aurora. Market sizing and forecasting is given for autonomous vehicle software licensing, as well the significant market potential for recurring revenue streams from essential and functional updates to autonomous vehicles over the course of their lifetime.

In a time of market consolidation, with many OEMs rationalizing their spend on autonomous vehicles and with many robotaxi startups beginning to feel the strain, this report can help guide OEMs to autonomous software development partners that can best help meet their autonomous and driverless objectives. At the same time, the report highlights which software development techniques and software tools can help autonomous software developers to secure vital revenue in the short term and position themselves effectively in the nascent autonomous vehicle market.

Table of Contents

  • 1. EXECUTIVE SUMMARY
    • 1.1. Introduction
    • 1.2. Artificial Intelligence (AI) Techniques Dominate Perception
    • 1.3. Deterministic Software Development Complements AI
    • 1.4. Software Toold Provide Short-Term Revenue Flow
    • 1.5. Expect Vendor Consolidation during the Coming Years
    • 1.6. End-to-End Stacks Will Triumph over Modular Approaches in the Short Term
    • 1.7. Recurring Revenue Streams
  • 2. OVERVIEW OF THE AUTONOMOUS DRIVING STACK
  • 3. PERCEPTION
    • 3.1. Vision-First Dominance
    • 3.2. Identification Using Convolutional Neural Networks
    • 3.3. Prediction Using Recurrent Neural Networks
    • 3.4. Conclusions
  • 4. SENSOR FUSION
    • 4.1. Early versus Late Sensor Fusion
  • 5. LOCALIZATION AND ENVIRONMENT MODELING
    • 5.1. Digital Maps in Autonomous Driving
    • 5.2. Relevant Use Cases for HD Maps
  • 6. MOTION PLANNING AND CONTROL
    • 6.1. A Multi-Agent Problem
    • 6.2. Behavior Planning
    • 6.3. End-to-End Deep Learning
    • 6.4. Deterministic Safety Monitors
    • 6.5. Mixed Deployment Scenarios
  • 7. AUTONOMOUS VEHICLE SOFTWARE DEVELOPMENT TOOLS
  • 8. AV SOFTWARE BUSINESS MODELS
    • 8.1. Licensing/Subscription Business Model
    • 8.2. End-to-End versus Modular Approaches
  • 9. AUTONOMOUS SOFTWARE VENDORS
    • 9.1. AImotive
    • 9.2. Aurora
    • 9.3. Elektrobit
    • 9.4. FiveAI
    • 9.5. Mobileye (An Intel Subsidiary)
    • 9.6. NVIDIA
    • 9.7. Zenuity
  • 10. MARKET EXPECTATION AND FORECASTS
    • 10.1. AV Software Licenses
    • 10.2. Recurring Revenue Opportunity
    • 10.3. Total Market Opportunity