Experiment Config

In the example you have learned how easy it is to run different experiment setups in combination with different planners. LocalPlannerBench is especially powerful because of the versatility of the supported experiment options and environments. To create your own custom experiment setup, simply write an exp.yaml file with your desired configurations. We will list the available options here:

T :int

Maximum simulated time to let the experiment run for.

dt :float

Timestep of the simulator.

dynamic :bool

Does the experiment contain dynamic obstacles or goals.

env :str

Name of the simulation environment. LocalPlannerBench currently supports 2 different simulators with a large number of environments and robots of varying fidelity from point mass robots to full fledged mobile manipulators. More information about the two simulators can be found here:

• gym_envs_planar

• gym_envs_urdf

goal :Dict

Dictionary containing a collection of subgoals for the current experiment. The keys of the subgoals should follow this structure: subgoal0, subgoal1, ..., subgoali. A subgoal specification is again very versatile, you can specify the indices of the joints, desired position, weight of the goal and even minimum and maximum goal values for random generation of goals. See motion_planning_scenes for more information.

initState :Dict

Should contain a dictionary with the initial state of the joints. The key q0 containing the initial positions and q0dot the initial velocities.

limits :Dict

Should contain a dictionary with the lower and upper limits of the states of the joints. The key low containing the lower bound positions and high the upper bound positions.

n :int

Number of joints

obstacles :Dict

Dictionary containing a collection of obstacles for the current experiment. The keys of the obstacles should follow this structure: obst0, obst1, ..., obsti. A subgoal specification is again very versatile, you can specify the dimension, desired position, and even minimum and maximum values for random generation of obstacles. See motion_planning_scenes for more information.

r_body :float

Radius of the links of the robot, to be used by planners for collision avoidance. Note: only a single radius for all links is currently supported.

randomObstacles :Dict

Dictionary containing information about the random generation of obstacles for this experiment. The Dictionary should contain the key number to specify the number of randomobstacles.

robot_type :str

The name of the type of robot used in the environment.

selfCollision :Dict

Dictionary containing information about configuration for self collision avoidance. The Dictionary should contain the key pairs to specify pairs of links that should avoid eachother.