To achieve commercialization of autonomous driving, enormous amount of real road testing is required for reliability validation. Real road tests on public roads are costly, and it is difficult to reconstruct extreme scenarios, and unable to cover comprehensive scenarios. This can be remedied by simulation tests which rebuild virtual scenarios and validate autonomous driving system, thus it can supplement the amount of data and achieve low cost.
At DeepRoute.ai, engineers input varying test scenarios like map, visual obstruction, jaywalking, emergency braking etc. Based on real road tests, and with data and parameter adjustment, a simulated scenario is constructed to encompass millions extreme cases. With a large amount of low-cost and highly efficient simulation training, the autonomous driving system keeps improving in its reliability, safety, and ability for generalization.
Alibaba Cloud provides elastic public cloud computing power for DeepRoute.ai. Through Elastic Container Instance (ECI) and Alibaba Cloud Container Service for Kubernetes (ACK), elastic scaling at second-level is achieved. A large amount of computing power can be allocated instantly, and matching ECS specifications can be selected based on business type. Once the task is completed, the computing power is released to save resources.
With the support of Alibaba Cloud’s elastic computing power,DeepRoute.ai simulation tests can be done at only 1% of what real road tests cost. Mileage accumulated in one month on public road tests can be finished in one day in simulation. Reinforced by both virtual and real road tests and validation, DeepRoute.ai Robotaxis can maneuver with confidence and efficiency on congested urban roads, narrow lanes and highways, dealing with different road conditions with safety in mind.
DeepRoute.ai Robotaxi operation spans across 200 kilometers with over 200 pick-up/drop-off points, covering commercial, residential, leisure areas and scenic spots. To date, DeepRoute.ai Robotaxi fleet has completed over 80,000 passenger trips, accumulating over 10 million kilometers.