Claims:WE CLAIM:
1. A method of avoiding collision between inspection devices (101) during inspection of an area, the method comprising:
receiving, by a collision avoidance system (105), spatial information (211) of the area from each of a plurality of inspection devices (101), initialized for inspecting the area, wherein each of the plurality of inspection devices (101) navigate across a plurality of hops and the spatial information (211) is received at beginning of each of the plurality of hops;
creating, by the collision avoidance system (105), a current local map (213) for each of the plurality of inspection devices (101) based on the spatial information (211) of a current hop of the plurality of hops of each of the plurality of inspection devices (101);
generating, by the collision avoidance system (105), a current global map (217) of the area by merging the current local map (213) of each of the plurality of inspection devices (101) with a previous global map (215) of the area, wherein the previous global map (215) corresponds to previous hops of the plurality of hops of each of the plurality of inspection devices (101);
localizing, by the collision avoidance system (105), each of the plurality of inspection devices (101) within the current global map (217) for determining a current location of each of the plurality of inspection devices (101);
estimating, by the collision avoidance system (105), one or more sub-paths (219) for navigating each of the plurality of inspection devices (101) from the current hop to a subsequent hop of the plurality of hops, by analysing the current global map (217) and the current location of each of the plurality of inspection devices (101) using a pre-trained path estimation model (229), wherein each of the one or more sub-paths (219) are validated for eliminating possibility of collisions among the plurality of inspection devices (101) in the subsequent hop; and
providing, by the collision avoidance system (105), each of the one or more sub-paths (219) to each of the plurality of inspection devices (101) for avoiding the collision between the plurality of inspection devices (101) during inspection of the area.

2. The method as claimed in claim 1, wherein the spatial information (211) is collected using one or more sensors (307) and image capturing devices (309) configured with each of the plurality of inspection devices (101).

3. The method as claimed in claim 1, wherein generating the current global map (217) of the area comprises:
identifying one or more adjacent local maps based on comparison among the current local map (213) corresponding to each of the plurality of inspection devices (101);
assigning a continuity score for each of the one or more adjacent local maps based on visual similarity among the one or more adjacent local maps; and
merging the one or more adjacent local maps into the current global map (217), when the continuity score of the one or more adjacent local maps is greater than a threshold score.

4. The method as claimed in claim 1, wherein validating each of the one or more sub-paths (219) comprises:
estimating a future location of the plurality of inspection devices (101) based on analysis of the current location of the plurality of inspection devices (101) and the one or more sub-paths (219);
determining possibility of collision between the plurality of inspection devices (101) at the future location of the plurality of inspection devices (101), using a pre-trained reinforcement model; and
eliminating the one or more sub-paths (219) having the possibility of collision.

5. The method as claimed in claim 1, wherein inspecting the area comprises:
capturing, using the plurality of inspection devices (101), a plurality of images of a region to be inspected within the area;
ranking each of the plurality of images based on one or more objects present in the plurality of images, wherein the ranking is performed using a pre-trained machine learning model;
combining one or more of the plurality of images into one or more inspection images based on the ranking; and
displaying the one or more inspection images on a display unit (206) associated with the collision avoidance system (105) for inspecting the area.

6. A collision avoidance system (105) for avoiding collision between inspection devices (101) during inspection of an area, the collision avoidance system (105) comprising:
a processor (203); and
a memory (205), communicatively coupled to the processor (203), wherein the memory (205) stores processor-executable instructions, which on execution, cause the processor (203) to:
receive spatial information (211) of the area from each of a plurality of inspection devices (101), initialized for inspecting the area, wherein each of the plurality of inspection devices (101) navigate across a plurality of hops and the spatial information (211) is received at beginning of each of the plurality of hops;
create a current local map (213) for each of the plurality of inspection devices (101) based on the spatial information (211) of a current hop of the plurality of hops of the each of the plurality of inspection devices (101);
generate a current global map (217) of the area by merging the current local map (213) of each of the plurality of inspection devices (101) with a previous global map (215) of the area, wherein the previous global map (215) corresponds to previous hops of the plurality of hops of each of the plurality of inspection devices (101);
localize each of the plurality of inspection devices (101) within the current global map (217) for determining a current location of each of the plurality of inspection devices (101);
estimate one or more sub-paths (219) for navigating each of the plurality of inspection devices (101) from the current hop to a subsequent hop of the plurality of hops, by analysing the current global map (217) and the current location of each of the plurality of inspection devices (101) using a pre-trained path estimation model (229), wherein each of the one or more sub-paths (219) are validated for eliminating possibility of collisions among the plurality of inspection devices (101) in the subsequent hop; and
provide each of the one or more sub-paths (219) to each of the plurality of inspection devices (101) for avoiding the collision between the plurality of inspection devices (101) during inspection of the area.

7. The collision avoidance system (105) as claimed in claim 6, wherein each of the plurality of inspection devices (101) collect the spatial information (211) using one or more sensors (307) and image capturing devices (309) configured with each of the plurality of inspection devices (101).

8. The collision avoidance system (105) as claimed in claim 6, wherein to generate the current global map (217) of the area, the processor (203) is configured to:
identify one or more adjacent local maps based on comparison among the current local map (213) corresponding to each of the plurality of inspection devices (101);
assign a continuity score for each of the one or more adjacent local maps based on visual similarity among the one or more adjacent local maps; and
merge the one or more adjacent local maps into the current global map (217), when the continuity score of the one or more adjacent local maps is greater than a threshold score.

9. The collision avoidance system (105) as claimed in claim 6, wherein to validate each of the one or more sub-paths (219), the processor (203) is configured to:
estimate a future location of the plurality of inspection devices (101) based on analysis of the current location of the plurality of inspection devices (101) and the one or more sub-paths (219);
determine possibility of collision between the plurality of inspection devices (101) at the future location of the plurality of inspection devices (101), using a pre-trained reinforcement model; and
eliminate the one or more sub-paths (219) having the possibility of collision.

10. The collision avoidance system (105) as claimed in claim 6, wherein to inspect the area, the processor (203) is configured to:
capture, using the plurality of inspection devices (101), a plurality of images of a region to be inspected within the area;
rank each of the plurality of images based on one or more objects present in the plurality of images, wherein the ranking is performed using a pre-trained machine learning model;
combine one or more of the plurality of images into one or more inspection images based on the ranking; and
display the one or more inspection images on a display unit (206) associated with the collision avoidance system (105) for inspecting the area.

Dated this 29th day of March 2019

MADHUSUDAN S.T
OF K&S PARTNERS
ATTORNEY FOR THE APPLICANT
IN/PA-1297
, Description:TECHNICAL FIELD
The present subject matter is, in general, related to automated inspection and more particularly, but not exclusively, to method and system for avoiding collision between inspection devices during inspection of an area.