Perception assistance

1 . A system for assisting an autonomous vehicle (AV), the system comprising:

a processor and non-transitory memory storing instructions that, when executed by the processor, cause the processor to perform remote assistance operations (RAOs);

the processor communicatively coupled to the AV; and

a user interface coupled to the processor;

wherein the RAOs are configured to perform an analysis of an obstruction and a current state of the AV and, based on a result of the analysis, present to a human Remote Advisor through the user interface perception adjustment options that are derived by simulating AV motion under varied perception parameters and by determining an adjustment that removes a predicted interaction between an AV footprint and an object footprint while maintaining AV safeguards;

wherein the user interface highlights areas of the predicted interaction;

wherein the presented perception adjustment options are parameter modifications to perception used by the AV to represent objects and plan motion during autonomous navigation and are distinct from selectable object classification labels and distinct from path or reroute commands; and

wherein the RAOs are further configured to perform one or more of:

accept a selection of a perception adjustment;

instruct the AV to adjust a perception of an object;

instruct the AV to adjust the AV footprint; and

instruct the AV to navigate autonomously after implementing the instructed adjustment.

2 . The system of claim 1 , wherein:

the object has an object footprint and has been assigned a class; and

the adjustment comprises:

an adjustment of the object footprint;

an adjustment of an attribute of the assigned class;

an allowed amount of intrusion of the object footprint into the AV footprint; or a combination thereof.

3 . The system of claim 2 , wherein the adjustment of the AV footprint comprises a reduction in a lateral buffer of the AV.

4 . The system of claim 2 , wherein:

a default AV footprint is stored in the AV and used in the autonomous navigation;

the adjustment of the AV footprint is performed on the stored default AV footprint; and the adjustment of the AV footprint is ended after the AV has moved past the object, thereby restoring use of the default AV footprint.

5 . The system of claim 2 , wherein:

the AV perceives the object by executing steps:

detecting an object based on a sensor output;

classifying the object as one of a predetermined set of classes; and

assigning a predetermined object footprint to the object based in part on the class of the object; and

the adjustment of the object footprint is performed on the assigned predetermined object footprint.

6 . The system of claim 5 , wherein the RAOs comprise a first restriction preventing the human Remote Advisor from selecting an adjustment of the perception of the object for one or more of the predetermined set of classes.

7 . The system of claim 5 , wherein the RAOs comprise a second restriction preventing the human Remote Advisor from implementing a perception override for one or more of the predetermined set of classes.

8 . A method of allowing adjustment of a perception of an object by an autonomous vehicle (AV), the method comprising:

providing a view of an environment around the AV through a sensor system of the AV to a human Remote Advisor;

performing an analysis of an obstruction and a current state of the AV and, based on a result of the analysis, presenting to the human Remote Advisor through a user interface perception adjustment options that are derived by simulating AV motion under varied perception parameters and by determining an adjustment that removes a predicted interaction between an AV footprint and an object footprint while maintaining AV safeguards;

highlighting, in the user interface, areas of the predicted interaction;

wherein the presented perception adjustment options are parameter modifications to perception used by the AV to represent objects and plan motion during autonomous navigation and are distinct from selectable object classification labels and distinct from path or reroute commands;

accepting, from the human Remote Advisor via the user interface, a selection of one of the presented perception adjustment options; and

instructing the AV to implement the selected perception adjustment and then navigate autonomously after implementing the selected perception adjustment.

9 . The method of claim 8 , wherein:

the AV has an AV footprint;

the object has an object footprint and has been assigned a class;

the perception adjustment is selected from a group of:

an adjustment of the AV footprint;

an adjustment of the object footprint;

an adjustment of an attribute of the assigned class; and

an allowed amount of intrusion of the object footprint into the AV footprint.

10 . The method of claim 9 , wherein the adjustment of the AV footprint comprises a reduction in a lateral buffer of the AV.

11 . The method of claim 9 , wherein:

a default AV footprint is stored in the AV and used in the autonomous navigation;

the adjustment of the AV footprint is performed on the stored default AV footprint; and the adjustment of the AV footprint is cancelled after the AV has moved past the object, thereby restoring use of the default AV footprint.

12 . The method of claim 9 , wherein:

the AV perceives the object by executing steps:

detecting an object based on a sensor output;

classifying the object as one of a predetermined set of classes; and

assigning a predetermined object footprint to the object based in part on the class of the object; and

the adjustment of the object footprint is performed on the assigned predetermined object footprint.

13 . The method of claim 12 , further comprising:

preventing the human Remote Advisor from selecting an adjustment of the perception of the object for one or more of the predetermined set of classes.

14 . The method of claim 12 , further comprising:

preventing the human Remote Advisor from selecting a perception override for one or more of the predetermined set of classes.

15 . A non-transitory memory contains instructions that, when loaded into a processor of a remote assistance platform (RAP) communicatively coupled to an autonomous vehicle (AV) and executed, cause the processor to cause the system to:

provide a view of an environment around the AV through a sensor system of the AV to a human Remote Advisor;

perform an analysis of an obstruction and a current state of the AV and, based on a result of the analysis, present to the human Remote Advisor through a user interface perception adjustment options that are derived by simulating AV motion under varied perception parameters and by determining an adjustment that removes a predicted interaction between an AV footprint and an object footprint while maintaining AV safeguards;

highlight, in the user interface, areas of the predicted interaction;

wherein the presented perception adjustment options are parameter modifications to perception used by the AV to represent objects and plan motion during autonomous navigation and are distinct from selectable object classification labels and distinct from path or reroute commands;

accept, from the human Remote Advisor via the user interface, a selection of one of the presented perception adjustment options; and

instruct the AV to implement the selected perception adjustment and then navigate autonomously after implementing the selected perception adjustment.

16 . The non-transitory memory of claim 15 , wherein:

the AV has an AV footprint;

the object has an object footprint and has been assigned a class;

the perception adjustment is selected from a group of:

an adjustment of the AV footprint;

an adjustment of the object footprint;

an adjustment of an attribute of the assigned class; and

an allowed amount of intrusion of the object footprint into the AV footprint.

17 . The non-transitory memory of claim 16 , wherein the adjustment of the AV footprint comprises a reduction in a lateral buffer of the AV.

18 . The non-transitory memory of claim 16 , wherein:

a default AV footprint is stored in the AV and used in the autonomous navigation;

the adjustment of the AV footprint is performed on the stored default AV footprint; and the adjustment of the AV footprint is cancelled after the AV has moved past the object, thereby restoring use of the default AV footprint.

19 . The non-transitory memory of claim 16 wherein:

the AV perceives the object by executing steps:

detecting an object based on a sensor output;

classifying the object as one of a predetermined set of classes; and

assigning a predetermined object footprint to the object based in part on the class of the object; and

the adjustment of the object footprint is performed on the assigned predetermined object footprint.

20 . The non-transitory memory of claim 19 , wherein the instructions further cause the processor to cause the non-transitory memory to perform one or more of:

prevent the human Remote Advisor from selecting an adjustment of the perception of the object for one or more of the predetermined set of classes; and

prevent the human Remote Advisor from selecting a perception override for one or more of the predetermined set of classes.