Announcing Applied’s Re-Simulation Tool Logstream

The log re-simulation tool Logstream enables AV and ADAS engineering teams to analyze disengagement events, evaluate stack performance, and bring safe autonomous systems to market faster.
Sep 21, 2021

Today, Applied Intuition is excited to announce Logstream, a re-simulation (closed-loop log replay) tool that allows autonomous vehicle (AV) and advanced driver-assistance systems (ADAS) engineering teams to analyze disengagement events, evaluate stack performance, and bring safe autonomous systems to market faster.

Logstream: Drive Log Re-Simulation

As automated driving systems need to safely navigate the real world, their testing and simulation environments need to be representative of the real world, including potential encounters of rare and dangerous events and on-road issues. Logstream is a log re-simulation tool that allows AV and ADAS engineers to analyze on-road test data and combine it with simulation to explore not only what happened but also what might have happened if the safety driver hadn’t intervened or another vehicle had acted differently. Our customers have already been using Logstream, and we are excited to share it with the broader AV ecosystem.

Logstream enables simulation engineers, motion planning engineers, and perception engineers to:

  • Automatically analyze disengagements to distinguish between safety-critical and unnecessary takeovers with confidence.
  • Use closed-loop re-simulation to validate scenarios beyond the point of disengagement by accurately modeling stack and vehicle behavior.
  • Analyze perception and localization performance by comparing stack outputs to manually labeled ground truth (Figure 1).
  • Turn one on-road issue into a full suite of test cases that fully evaluates the broader scenario by augmenting the logged data.
  • Run regression tests to validate new software versions and prevent re-introducing on-road issues.
Figure 1: Logstream allows perception engineers to analyze perception and localization performance by comparing stack outputs to manually labeled ground truth.

Contact our engineering team if you are interested in seeing a Logstream product demo.