CV2 Error about lqrrt HOT 6 CLOSED

Yes, as suspected the problem is that somehow ss_img is getting one dimension of length zero. Based on the 0's in push's y-components, I think I know exactly why this is happening. Will fix soon.

from lqrrt.

My current hypothesis is that one of the dimensions of ss_img is going to zero length, but based on the way things were when this happened, I'm not sure why that'd be the case. I am going to surround that block with a try-except so we can get a look at ss_img (and every argument of that floodFill) if this ever happens again.

test_flood = np.copy(ss_img)
try:
    area, rect = cv2.floodFill(test_flood, np.zeros((test_flood.shape[0]+2, test_flood.shape[1]+2), np.uint8), ss_goal, 69)
    if test_flood[ss_seed[1], ss_seed[0]] == 69:
        gs[:2] = (np.add([col, row], [last_offsets[0], last_offsets[1]]).astype(np.float64) / self.ogrid_cpm) + self.ogrid_origin
        found_entry = True
        break
except:
    print("\nA sample space image during select_exploration may have become degenerate. How odd.")
    print("type(ss_img): {}".format(type(ss_img)))
    print("np.shape(ss_img): {}".format(np.shape(ss_img)))
    print("current push: {}".format(push))
    print("ss_goal: {}".format(ss_goal))
    print("\nTerminating.")
    self.failure_reason = "glitch"
    self.set_goal(self.state)
    return(1, escape.gen_ss(self.next_seed, self.goal), np.copy(self.goal))

from lqrrt.

There I was minding my own business driving around and all of a sudden this happened to me!

Preparing: drive 
Rotating towards goal (duration: 1.36)
OpenCV Error: Bad flag (parameter or structure field) (Unrecognized or unsupported array type) in cvGetMat, file /home/matt/Downloads/opencv-2.4.9/modules/core/src/array.cpp, line 2482

A sample space image during select_exploration may have become degenerate. How odd.
type(ss_img): <type 'numpy.ndarray'>
np.shape(ss_img): (601, 0)
current push: [80, 40, 0, 0]
ss_goal: (120, 40)

Terminating.

np.shape(ss_img): (601, 0) << Could this be an issue?

I set the waypoint towards the bottom left corner of this grid. It had started rotating then after a second or two it gave the error and terminated the move.

from lqrrt.

Also, in boundary_analysis I'm encountering the same issue (with it's floodFill)

from lqrrt.

The current fix just picks some basic sample space (unadvised by the ogrid geometry, like the old days) when we get into the special (and rare) degenerate case. I haven't been able to recreate the degenerate case since implementing this fix, so it may not work as intended, and thus this issue will not close yet. Please report here if the print statement tells you to.

from lqrrt.

It seems that there are multiple causes of the degenerate case. One we know for sure is being too close to the edge of the ogrid. Regardless, the resulting failure is the same in every case, so we have protected against it by catching the degenerate ss_img and reverting to the old style ogrid analysis as needed.

from lqrrt.