US 12,321,178 B2
Semantic models for robot autonomy on dynamic sites
Marco da Silva, Waltham, MA (US); Dom Jonak, Winchester, MA (US); Matthew Klingensmith, Somerville, MA (US); and Samuel Seifert, Waltham, MA (US)
Assigned to Boston Dynamics, Inc., Waltham, MA (US)
Filed by Boston Dynamics, Inc., Waltham, MA (US)
Filed on Jan. 26, 2022, as Appl. No. 17/648,942.
Claims priority of provisional application 63/143,528, filed on Jan. 29, 2021.
Prior Publication US 2022/0244741 A1, Aug. 4, 2022
Int. Cl. G05D 1/00 (2024.01); B62D 57/032 (2006.01); G01C 21/00 (2006.01); G06V 20/50 (2022.01)
CPC G05D 1/0274 (2013.01) [B62D 57/032 (2013.01); G01C 21/383 (2020.08); G05D 1/0214 (2013.01); G05D 1/0276 (2013.01); G06V 20/50 (2022.01)] 24 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving, by data processing hardware of a robot, first sensor data from one or more sensors of the robot, the first sensor data corresponding to an environment;
receiving, by the data processing hardware, a semantic model associated with the environment, the semantic model comprising semantic information identifying one or more objects located within the environment;
identifying, by the data processing hardware, a first feature within the environment based on the first sensor data;
determining, by the data processing hardware, that the first feature as identified based on the first sensor data corresponds to an object of the one or more objects located within the environment and identified by the semantic information of the semantic model;
based on determining that the first feature as identified based on the first sensor data corresponds to the object, generating, by the data processing hardware, in a first map, a first localization reference point corresponding to the first feature;
generating, by the data processing hardware, in a second map, a first identifier of a no-step region or an obstacle corresponding to the first feature as identified based on the first sensor data, wherein the first sensor data, the first map, and the second map correspond to different features;
instructing, by the data processing hardware, localization of the robot within the environment using the first map, wherein instructing the localization of the robot comprises:
instructing, by the data processing hardware, identification of a subset of features as identified based on second sensor data that correspond to the first map, and
based on generating the first localization reference point in the first map and the identification of the subset of features as identified based on the second sensor data that correspond to the first map, instructing, by the data processing hardware, determination of a location of the robot relative to the first feature as identified based on the second sensor data; and
instructing, by the data processing hardware, performance of an action by the robot based on the localization of the robot and using the second map.