NVIDIA DriveWorks SDK is the foundation for autonomous vehicle (AV) software development, providing an automotive-grade middleware with accelerated algorithms and versatile tools. DriveWorks enables developers to innovate their own AV solutions by offering modular, open, and customizable components. Developers can use individual modules for specific functions or combine multiple modules to achieve higher-level objectives.

Features

• Compute Graph Framework (CGF) and STM Scheduler: Express applications as nodes of a directed acyclic graph (DAG) with fine-grained control over compute resource consumption. The STM scheduler ensures efficient, non-preemptive scheduling for safety-critical applications.
• Sensor Abstraction Layer (SAL): Provides a unified interface to vehicle sensors, synchronized timestamping, data serialization, and replay capabilities. Supports a diverse set of Hyperion-compatible sensors and custom sensor integration.
• Image and Point Cloud Processing: Offers optimized modules for feature detection, tracking, structure from motion, rectification, lidar packet accumulation, registration, and planar segmentation.
• VehicleIO: Supports multiple production drive-by-wire backends and custom interface integration through the VehicleIO Plugin framework.
• DNN Framework: Enables loading and inferring TensorRT models with custom layer support and inference acceleration using DRIVE AGX’s integrated GPU.
• Recorder: Tools for recording, synchronizing, and playing back sensor data for training and development.
• Calibration: Provides tools for static and self-calibration of AV sensors, ensuring high fidelity and real-time compensation for environmental changes.
• Egomotion: Uses motion models to track and predict vehicle pose, supporting odometry-only and IMU-based models.

Benefits

• Accelerated Development: Quickly build and deploy AV solutions with high-performance, real-time processing capabilities.
• Enhanced Efficiency: Improve operational efficiencies with optimized algorithms and reliable performance.
• Scalability: Suitable for large-scale deployments, ensuring reliable performance in enterprise environments.
• Flexibility: Supports a wide range of sensors and deployment scenarios, from local servers to cloud environments.
• Comprehensive Support: Benefit from extensive tools and frameworks for streamlined development and integration.