Automotive

Real-world miles aren't enough

Autonomy programs depend on road tests that miss rare, mission-critical events. Putting schedules, and safety at risk.

automotive_compilation from Parallel Domain on Vimeo

Playing in picture-in-picture

Software augmented testing grounded in your drive logs.

Real-fidelity neural reconstructions and deterministic simulation to confidently test and validate ADAS and AV systems continuously through development.

Challenges with AV and ADAS Perception Development

Safe, Reliable Perception

Edge Cases – Unique scenarios are critical for model accuracy and safety, but rarely seen in the real-world and often unsafe or impossible to capture.

Observability – Changes to a perception model can have unintended consequences that are not observable until tested in the real-world.

Safety – Crashes can lead to litigation, regulatory scrutiny, and diminish brand perception.

Solution – Continuously evaluate both standard and edge-case scenarios in a virtual world designed to mimic the real-world

Time and Cost

Collection – It takes months to collect, label, and QA real sensor data.

Evaluation – It takes weeks to test in the real-world.

Staffing a team to capture, label, curate, QA, and field-test is expensive.

Solution – Generate the scenarios, test suites, and datasets you need in days, not months

Scalability Across ODD Domains and Sensor Configurations

New vehicle rigs and sensor types require recapturing sensor data, labeling, curation, and QA.

Entering new geographic markets requires new data and overcoming regulatory hurdles.

Regulation is constantly changing and evolving requiring rework to perception models.

Solution – Simulate new camera configurations, environments, and scenarios by updating code, not recapturing data

PD Replica + PD Sim for ADAS and AV perception validation.

Evaluate

Open-loop and closed-loop testing integration for automotive perception stacks. Nightly regression testing for lane-keeping, object detection, and path planning. Perception unit testing across weather, lighting, and traffic density. Near-validation testing in real-world scanned environments (PD Replica).

Analyze

Benchmark against highway, urban, and suburban scenarios with configurable sensor suites. Explore sensor trade-offs across lidar, camera, and radar configurations. Prove safety cases across ODD boundaries using accident reconstructions and regulatory scenarios.

Train

Scaled data generation for detection, segmentation, and tracking model training. Infinite road scene variations across weather, lighting, and traffic conditions. Simulate in real-world location scans for domain-specific training data (PD Replica).

PD Replica - Closing the sim-to-real gap with real environments

Incorporate real-world scans as fully annotated, simulation-ready environments seamlessly integrated into Parallel Domain’s Data Lab API. Experience unparalleled variety and realism for model testing, training, and validation.

Addressing Industry Challenges

Accelerate validation without fleet road testing

Compared to real-world testing, simulations can be run in parallel, enabling testing in hours what would take weeks to test on the road. For training use-cases, generate annotated sensor data in weeks rather than months for capturing, labeling, curating, and QAing real-world data.

Scale across ODD domains and sensor stacks

Simulation assets and environments can be updated quickly to match new regional needs. No need to recapture data when hardware changes, it is just a few small tweaks to our simulation sensor configurations.

Prove safety for passengers and vulnerable road users

Ensuring safety and reliability comes from routinely testing your AI system against critical, rare, and exhaustive scenarios. Doing this across thousands of variations, in real (PD Replica) or procedural environments provides additional assurance of systems performing as designed.