Autonomous driving (AD) systems rely on sophisticated software stacks to process sensor data, make decisions, and control vehicles. These stacks typically incorporate complex sensing suites, covering a wide range of scenarios to ensure safety and reliability. However, as we transition into the initial phases of passenger-vehicle AD integration, which is beginning with lower autonomy levels, there’s an evident need for a smaller sensing suite.

It’s essential to address the challenge of adapting software stacks that were initially implemented in robotaxis and commercial vehicles, for smaller sensing suites. As varying levels of autonomy are deployed across a wider variety of vehicles, this creates the need for smaller sensing suites in mass-market vehicles with lower autonomy levels, such as L2+; while higher-end vehicles with fully autonomous modes require comprehensive sensor coverage.

This presentation outlines how an AD stack can be efficiently scaled down to L2+, and simultaneously ported to a production chip.