Racetrack Domain

The racetrack domain is a standard testbed problem often used in Reinforcement Learning. This page presents an adapted version that has been used as a testbed for single-agent search problems. The domain is unique because the edges are directed and there is a goal-test function instead of a single goal state. The problem definition from Sutton and Barto (2018) is as follows:
Consider driving a race car around a [track]. You want to go as fast as possible, but not so fast as to run off the track. In our simplified racetrack, the car is at one of a discrete set of grid positions [...]. The velocity is also discrete, a number of grid cells moved horizontally and vertically per time step. The actions are increments to the velocity components. Each may be changed by +1, −1, or 0 in each step, for a total of nine (3 × 3) actions. Both velocity components are restricted to be nonnegative and less than 5, and they cannot both be zero except at the starting line.
An implementation of the racetrack domain is below. The racecar starts on the top row heading right and must reach any cell on the bottom row. Try to solve a problem as efficiently as possible and compare to the optimal solution.

Instructions

  1. Reset the problem or choose a map to begin.
  2. Your racecar is drawn as a black triangle.
  3. Your possible moves are drawn in yellow.
  4. Click on a yellow cell to advance your racecar to that location.
  5. Reach the bottom row as quickly as possible.

Control:

Change Map:

Research Citations