Autonomous Planning and Decision Making

What It Is

Describes the methods used by autonomous driving systems to make and execute navigation decisions based on sensor data.

  • Planning refers to the ability to make autonomous decisions based on fused sensor data consistently with the goals of the driving system. Decision making is the execution of such decisions.  

    The autonomous driving system engages in three qualitatively different forms of planning 

  • Mission planning includes the system's ability to plan its pathway to its final destination  
  • Behavioral planning includes the system's ability to react to signals and objects internal and external to the vehicle such as obstacles, traffic signals, and humans 
  • Motion planning includes the system's ability to plan its path trajectory, speed, and acceleration.

Machine vision refers to the system's perception of its surroundings based on fused data from its sensor, GPS, and location data. The system uses its perception of its surroundings to plan a course of action in pursuit of its predetermined goals. There are two main models of machine vision:

  • Under the object-based model, the system fuses sensor data from cameras, radars, ultrasonic, and lidars to recognize objects and form a full representation of its surroundings, as illustrated in the figure below. The state of the environment is continuously updated to be consistent with the most recently processed sensor data. The system then uses the latest constructed representation of its surroundings to make decisions. This paradigm provides the driving system with a more robust view of its environment but requires a longer processing period. This is also referred to as the mediated perception paradigm for autonomous driving. 

(Dietmayer, 2016)

  • Under the grid-based modelthe system directly maps a sensory input to a decision. Rather than maintain a full representation of its surroundings, the system divides its local environment into a grid of cells of equal size and assesses any obstacles within each cell as the vehicle moves, as illustrated in the figure below. The system directly uses its detection of obstacles in grid cells to make decisions. This paradigm provides the driving system with a less robust view of the its environment, but offers quicker processing period. This is also referred to as the behavior reflex paradigm for autonomous driving. 

(Dietmayer, 2016)

Explainer Editors
Michael Clamann, PhD, CHFP
Recommended Citation

Duke SciPol, “Autonomous Planning and Decision Making” available at​ (06/12/18).

Explainer Last Updated Date
Monday, April 23, 2018
Explainer Type
Emerging Tech

Creative Commons License This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. Please distribute widely but give credit to Duke SciPol, linking back to this page if possible.