You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: plannerarena/www/help.md
+4-5Lines changed: 4 additions & 5 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -224,7 +224,7 @@ The ompl_benchmark_statistics.py script can produce a series of plots from a dat
224
224
-*date:* the date on which the experiment was performed.
225
225
-*seed:* the random seed used.
226
226
-*setup:* a string containing a “print-out” of all the settings of the SimpleSetup object used during benchmarking.
227
-
227
+
228
228
Any additional columns are assumed to be numeric values corresponding to experimental hyperparameters. This can be useful to show planner performance as a function of, e.g., number of revolute joints for a parametric robot arm or number of obstacles in parametric environment. Planner Arena will show a selection widget for each hyperparameter. The user can choose to (1) aggregate planner runs over all hyperparameter values, (2) show performance separated out by hyperparameter value, or (3) show performance for one selected hyperparameter value.
229
229
-**plannerConfigs**. There are a number of planner types (such as PRM and RRT), but each planner can typically be configured with a number of parameters. A planner configuration refers to a planner type with specific parameter settings. The `plannerConfigs` table contains the following information:
230
230
-*id:* an ID used in the `runs` table to denote that a given planner configuration was used for a run.
@@ -238,8 +238,8 @@ The ompl_benchmark_statistics.py script can produce a series of plots from a dat
238
238
-*id:* ID of the run
239
239
-*experimentid:* ID of the experiment to which this run belonged.
240
240
-*plannerid:* ID of the planner configuration used for this run.
241
-
242
-
In addition, there will be many benchmark statistics. None are *required*, but the OMPL planners all report the properties described above such as time, memory, solution length, simplification time, etc. It is possible that not all planners report the same properties. In that case, planners that do not report such properties will have NULL values in the corresponding fields. Users can programmatically define new properties that can get logged for each run in OMPL.
241
+
242
+
In addition, there will be many benchmark statistics. None are *required*, but the OMPL planners all report the properties described above such as time, memory, solution length, simplification time, etc. It is possible that not all planners report the same properties. In that case, planners that do not report such properties will have NULL values in the corresponding fields. Users can programmatically define new properties that can get logged for each run in OMPL.
243
243
-**progress**. Some planners (such as RRT*) can also periodically report properties *during* a run. This can be useful to analyze the convergence or growth rate. The `progress` table contains the following information:
244
244
-*runid:* the ID of the run for which progress data was tracked.
245
245
-*time:* the time (in sec.) at which the property was measured.
@@ -248,6 +248,5 @@ The ompl_benchmark_statistics.py script can produce a series of plots from a dat
248
248
-*iterations:* the number of iterations.
249
249
-*collision_checks:* the number of collision checks (or, more precisely, the number state validator calls).
250
250
-*best_cost:* the cost of the best solution found so far.
251
-
251
+
252
252
As with run properties, users can programmatically define their own progress properties that will be logged during each run of a planner.
0 commit comments