JournalResearchFieldsDatabaseWorkspace
← Fields

Autonomous Driving

Autonomous-driving analysis follows one scenario from calibrated sensor signals to a world model, candidate futures, constrained motion, executed control, and a defined fallback when the primary path degrades.

scenariofallbackcandidate trajectories

Environment

Scenario and sensing

  • road
  • actor
  • sensor
  • map
  • occlusion

World model

Perception and prediction

  • object
  • lane
  • occupancy
  • future state

Motion

Planning and control

  • candidate
  • cost
  • trajectory
  • command

Operations

Service and safety

  • dispatcher
  • operator
  • fallback
  • rider state
ObjectStructureRecordForm
Scenario familyRoad geometry, actors, occlusion, traffic control and vehicleSpatial relation, condition, time and affected behaviorScenario map
Sensor processingSensor, frame, calibration, preprocessing and fusionModality, mounting, synchronization and coordinate conversionSensor chain
Perception outputDetected object, lane, occupancy cell and stateObject fields, confidence, spatial frame and updateSpatial object schema
PlanningCandidate trajectories, constraints, cost and selected trajectoryCandidate construction, objective, constraint and selectionTrajectory set
Control stackSensor subsystem, perception, prediction, planner and controllerLayer interface and data passed between layersLayered control stack
Fleet operationsVehicles, human operators, resources, dispatcher and tasksAssignment criteria, operating state and recovery pathFleet / state map
Rider interfaceInteraction states, events, workflow steps and system responseState transition and associated vehicle or service actionUI state series
Safety fallbackDegradation, safe action, remote operator and stated reasonDetection condition, risk state, fallback sequence and recoveryFallback decision map
Temporal and spatial evidenceTrajectories, velocity profiles, occupied intervals, image pyramidTime base, reference path, cost, scale and selected outputProfile / timeline / pyramid

space + actors + condition

Scenario record

The environment and event that cause a planning or control decision.

candidates → cost → selection

Trajectory record

Alternative paths, constraints, selected motion, and time-dependent profiles.

degradation → fallback → recovery

Safety record

Detected condition, safe action, human involvement, and exit state.

task → resource → state

Fleet record

How vehicle and human resources are assigned, monitored, and recovered.

Calibrate

Camera, lidar, radar, or map data is synchronized and placed in a common coordinate frame.

Model

Objects, lanes, occupancy, confidence, and future states form an internal scene representation.

Select

Candidate trajectories are tested against kinematic, comfort, and safety constraints and a stated cost.

Execute

The selected trajectory becomes steering, acceleration, braking, or other vehicle commands.

Degrade

A monitored fault transfers control, narrows the safety envelope, or initiates a minimal-risk maneuver.

The trace keeps nominal planning, physical control, and fallback evidence connected while preserving their different triggers and states.

Sensor signal ≠ semantic world objectScenario family ≠ single illustrationRider state ≠ fleet stateSafety result ≠ fallback mechanism

Why is a scenario better represented as a family than as one road image?

Variants can preserve the same road, actors, and identifiers while changing one constraint—occlusion, gap size, boundary type, signal state, or reachable stopping area—and showing the changed behavior.

  • shared scene elements
  • changed condition
  • affected decision
  • before / after behavior

How does a detected object become a vehicle trajectory?

Perception produces a world-model object; prediction adds candidate future states; planning maps constraints and costs to candidate trajectories; control converts the selected trajectory into vehicle commands.

  • object schema and frame
  • prediction horizon
  • constraint set and cost
  • selected command

What makes a fallback technically defined?

The record states the detected fault or degraded condition, the graded action, the handover or stopping constraint, and the state that permits recovery. Slowing, pulling over, and remote intervention are different branches.

  • fault or trigger
  • safety envelope
  • handover timing
  • recovery condition

A driving capability is described by two traceable paths: the nominal conversion of scene state into motion and the degraded conversion of a fault or uncertainty into a bounded fallback and recovery state.

  1. Scenario variants isolate which environmental condition changes the vehicle decision.

  2. Perception, prediction, planning, and control exchange different objects and should not be collapsed into one stack diagram.

  3. A fallback becomes defined when trigger, graded action, operating constraint, human or fleet involvement, and exit state are connected.

  • Scenario variants preserve shared elements and identify the changed condition.
  • Sensor, world-model, prediction, and trajectory records use aligned frames and time.
  • Fallback triggers, actions, and recovery criteria are stated at the same level of specificity.