Scale AI > Case Studies > Enhancing Autonomous Trucking with Synthetic Data: A Kodiak Robotics Case Study

Enhancing Autonomous Trucking with Synthetic Data: A Kodiak Robotics Case Study

Technology Category

Cybersecurity & Privacy - Identity & Authentication Management
Sensors - Autonomous Driving Sensors

Applicable Industries

Plastics
Transportation

Applicable Functions

Logistics & Transportation

Use Cases

Autonomous Transport Systems
Virtual Training

Services

System Integration
Training

About The Customer

Kodiak Robotics is a Mountain View, CA-based autonomous technology company that is revolutionizing the long-haul trucking industry. The company is developing self-driving capabilities and technologies to navigate all aspects of highway driving and deliver freight efficiently and on-time. The team, which includes several self-driving industry veterans, is committed to building the world's most efficient, reliable, and respected end-to-end delivery solution. Kodiak's unique approach involves leveraging a sensor fusion system combined with a lightweight mapping solution.

The Challenge

Kodiak Robotics, an autonomous technology company, is developing self-driving capabilities for the long-haul trucking industry. The company uses a unique sensor fusion system and a lightweight mapping solution to navigate highway driving and deliver freight efficiently. However, the company faced a significant challenge in training its software to handle rare scenarios, such as pedestrians walking on the highway. These edge cases are crucial for a production-level autonomous vehicle system, but collecting enough real-world examples to train the models reliably was proving difficult.

The Solution

To overcome this challenge, Kodiak Robotics partnered with Scale to provide synthetic data to augment their existing ground-truth training data with simulated pedestrians. Scale's unique human-in-the-loop synthetic data generation process creates diverse and realistic synthetic data. Trained taskers validate the placement and poses of synthetic pedestrians to ensure the synthetic data is realistic. The data is delivered via the same dashboard and APIs as Kodiak's existing annotation pipeline, ensuring seamless integration. Additionally, Kodiak uses Nucleus to identify specific scenes in their dataset that had edge cases they needed to improve their model on. This includes scenes where construction workers are present and where a vehicle is traveling under a bridge.

Operational Impact

The use of synthetic data has significantly improved Kodiak's ability to handle edge cases in autonomous trucking. By centralizing all their data, including multiple labeling projects and raw, unlabeled data, into a single dataset, the team can quickly iterate on model experiments, query for specific attributes or metadata on the fly, and close the loop for a more end-to-end data and model management system. Going forward, the team can review both insights and model metrics in Nucleus to identify scenes with poor IoU (intersection over union) and curate subsets of data where their model wasn't performing well, in which additional synthetic data might be helpful. This approach has enhanced the robustness of Kodiak's autonomous trucking model, making it more reliable and efficient.