I am planning with RRTstar for a 6 DOF articulated robot. I move from a joint start to a pose goal. I plot mean of path length vs planning time and do 10 runs for each 0.1s intervall in [0.1; 9.9]

Although I see high variability, the mean of the path length does not decrease when planning time increases. I am quite surprised by this since I have set the optimization objective to PathLengthOptimizationObjective. I see the same behaviour when I play with num_planning_attempts or range.

Is there something I am missing like enabling the planner params set by setPlannerParams()? Or is this expected behaviour since I am not planning around an obstacle, but rather "besides" it.

Here is my code:

move_group_GR.setPoseReferenceFrame("world_zero");
bool success = false;
int num_runs = 10;
double max_planning_time = 10.0;
double min_planning_time = 0.1; 
double planning_time_step = 0.1;

for(float a = min_planning_time; a <= max_planning_time; a = a + planning_time_step ){
    for(int i = 0; i < num_runs; i++){

    move_group_GR.setNumPlanningAttempts(10);
    move_group_GR.setPlanningTime(a);
    move_group_GR.setPlannerId("RRTstar");
    std::map<std::string, std::string> m;
    m.insert(std::pair<std::string, std::string>("range", "0.0"));
    m.insert(std::pair<std::string, std::string>("goal_bias", "0.05"));
    m.insert(std::pair<std::string, std::string>("delay_collision_checking", "1"));
    m.insert(std::pair<std::string, std::string>("optimization_objective", "PathLengthOptimizationObjective"));
    move_group_GR.setPlannerParams("RRTstar", "Gripper", m);

    // Target
    move_group_GR.setPoseTarget(t_pose);
    // Start
    move_group_GR.setStartStateToCurrentState();

    // Generate Plan
    moveit::planning_interface::MoveGroupInterface::Plan measurePlan;
    success = false;
    success = (move_group_GR.plan(measurePlan) == moveit::planning_interface::MoveItErrorCode::SUCCESS);

    // Executing and logging ...

    // Move back to start ...
    }
}