US 12,306,632 B2
Method and system for semantic navigation using spatial graph and trajectory history
Kritika Anand, Hyderabad (IN); Snehasis Banerjee, Kolkata (IN); Brojeshwar Bhowmick, Kolkata (IN); Madhava Krishna Krishnan, Hyderabad (IN); Gulshan Kumar, Hyderabad (IN); and Sai Shankar Narasimhan, Hyderabad (IN)
Assigned to Tata Consultancy Services Limited, Mumbai (IN)
Filed by Tata Consultancy Services Limited, Mumbai (IN)
Filed on Feb. 22, 2023, as Appl. No. 18/172,663.
Claims priority of application No. 202221009498 (IN), filed on Feb. 23, 2022.
Prior Publication US 2023/0266766 A1, Aug. 24, 2023
Int. Cl. G05D 1/00 (2024.01); G05D 1/02 (2020.01)
CPC G05D 1/0217 (2013.01) [G05D 1/0088 (2013.01); G05D 1/0246 (2013.01)] 9 Claims
OG exemplary drawing
 
1. A processor implemented method comprising:
receiving, by one or more hardware processors, a target object to be reached by a mobile robot in an indoor environment, wherein the indoor environment comprises a plurality of regions, wherein each of the plurality of regions is associated with a plurality of objects; and
iteratively performing till the mobile robot reaches the target object:
identifying, by the one or more hardware processors, a current location of the mobile robot based on a plurality of visible regions among the plurality of regions around the mobile robot using a localization technique;
computing, by the one or more hardware processors, an embedding corresponding to each of the plurality of visible regions based on the current location of the mobile robot, a plurality of valid trajectory paths of the mobile robot and the target object, using a pretrained Graph Network (GNN), wherein the GNN is pretrained using trajectory data and a spatial relationship graph associated with the indoor environment, wherein the spatial relationship graph with a plurality of nodes and a plurality of edges is generated based on a spatial co-occurrence statistics data of the plurality of regions and the associated plurality of objects, by:
generating the plurality of nodes by creating a node for each of the plurality of regions and each of the plurality of objects corresponding to each of the plurality of regions;
generating the plurality of edges by creating (i) an edge between each of the plurality of regions and the plurality of objects associated with each of the plurality of regions and (ii) an edge between each of the plurality of regions;
obtaining a spatial co-occurrence weight for each of the plurality of edges based on a frequency statistics data, wherein the spatial co-occurrence weight for each of the plurality of edges between each of the plurality of regions and each of the corresponding plurality of objects is obtained by:
obtaining a frequency of occurrence of each of the plurality of objects corresponding to each of the plurality of regions; and
computing the spatial co-occurrence weight by normalizing the frequency of occurrence of each of the plurality of objects based on a corresponding frequency of occurrence of the region associated with each of the plurality of objects; and
updating the obtained spatial co-occurrence weight for each of the plurality of edges of the spatial relationship graph;
computing, by the one or more hardware processors, a similarity score for each of the plurality of visible regions based on the corresponding embedding, using a scoring technique;
identifying, by the one or more hardware processors, an optimal visible region from the plurality of visible regions by comparing the similarity score corresponding to each of the plurality of visible regions, wherein the visible region with a maximum similarity score from among the plurality of visible regions is identified as the optimal visible region; and
selecting, by the one or more hardware processors, a next action to be taken by the mobile robot from a plurality of actions based on the optimal visible region, wherein the mobile robot moves towards the optimal visible region and checks whether it is able to view the target object after reaching the optimal visible region.