calc_score of candidate in bbs

Hi author,
In code we use DiscreteTransformation::calc_score() to score each candidate,

hdl_global_localization/src/hdl_global_localization/bbs/bbs_localization.cpp

Lines 59 to 61 in 3586396

    
           const auto& gridmap = gridmap_pyramid[level]; 
        
           auto transformed = transform(points, gridmap->grid_resolution(), theta_resolution); 
        
           score = gridmap->calc_score(transformed);

if we use the transformed points creates a grid_map with OccupancyGridMap::insert_points(), and then comapre the difference pixel by pixel, and then sum the whole difference value. We also modify bbs_localization.cpp#L27, to return score > rhs.score;. The smaller difference, the better the candidate.
Have you tried this scoring method? Thanks for your help and time !

This algorithm finds the transformation that maximizes the log likelihood (e.g., num of hit count), thus the larger score, the better. We can reform it as a minimization problem by modifying DiscreteTransformation::operator< as well as OccupancyGridMap::pyramid_up, but I suppose it doesn't make a significant difference on the branch pruning efficiency because it's just an inverse of the current formulation.

Thanks author :), my problem is solved.

	const auto& gridmap = gridmap_pyramid[level];
	auto transformed = transform(points, gridmap->grid_resolution(), theta_resolution);
	score = gridmap->calc_score(transformed);