Small decisions. Interesting possibilities.Submit contentSubmit

How it uses Jev

Each driving decision request summarizes speed, route, road boundaries, lane position, traffic, intersections, and candidate trajectories. Jev chooses whether to drive or stop, then selects a trajectory when driving; an optional route choice is added when multiple routes are available. Code expands candidate aliases and applies the selected controls in the simulator.

What Jev decides

Illustrative driving motion decisionExample answers · not a recorded Jev response · Source ↗
Question 1 · motion
YOUR APP
INSTRUCTION

Choose whether to drive or stop using current conflicts, legal requirements, and stop memory; proximity alone is not a reason to stop.

STATE

Representative simulator state with moving trajectory candidates and an eligible zero-speed stop candidate; the motion options are conditional on those candidates being available.

JEV · CHOICE
  1. drive
  2. stop

App workflow

  1. Collect simulator state

    Gather vehicle speed, route progress, roadway geometry, traffic, hazards, and candidate trajectories.

  2. Prepare a bounded request

    Compact the state, add currently available motion and trajectory questions, and alias candidate IDs.

  3. Choose motion and path

    Ask Jev to drive or stop, then select a movement candidate if driving; add a route choice only when alternatives exist.

  4. Apply simulator controls

    Expand aliases and pass the selected motion/path into deterministic vehicle control.

Why it is interesting

The request supplies bounded candidate paths and compact physical context instead of asking Jev to generate control values. Drive/stop is separated from the path choice so many similar movement candidates do not dilute the stop decision.