Determining error event likelihoods for vehicle routes and route paths
Systems and techniques are provided for determining error event likelihoods of routes of an autonomous vehicle (AV). An example method can determine, for each route from a plurality of routes, a respective set of route paths for the route, each route located within a geofence; determine, for each route path of each route, a first respective likelihood that the AV will use the route path for the route; determine, for each route path of each route, a second respective likelihood that the AV will encounter an error event along the route path during an autonomous operation; determine, for each route path of each route, an overall error event likelihood based on the first and second respective likelihoods for each route path of that route; and determine an aggregated error event likelihood for each route based on the overall error event likelihood of each route path of that route.
1 . A system for managing an autonomous vehicle (AV) comprising:
a memory; and
one or more processors coupled to the memory, the memory storing instructions that define a simulation framework configured to generate simulations of the AV using models of its real-world operation within a model of a real-world geofenced operational design domain (ODD), wherein the one or more processors are configured to:
generate a plurality of simulated routes for autonomous operation of the AV located within the ODD from a starting location to a destination location;
determine, for each simulated route from the plurality of simulated routes, respective route paths from the starting location of the route to the destination location of the route;
determine, for each route path of each simulated route, a first respective likelihood that the AV will use the route path for an AV trip comprising the route;
determine, for each route path of each simulated route, a second respective likelihood that the AV will encounter one or more error events along the route path during the AV trip;
determine, for each route path of each simulated route, an overall error event likelihood based on the first respective likelihood and the second respective likelihood determined for each route path of the simulated route;
determine an aggregated error event likelihood for each simulated route based on the overall error event likelihood of each route path of the simulated route;
determine an error event likelihood for the geofence based on the aggregated error event likelihood for each simulated route; and
determine an impact on the error event likelihood of the geofence by a simulation of at least one of a proposed change to a software of the AV, a sensing capability of the AV, one or more constraints of the AV, one or more parameters of an operating design domain (ODD) associated with the geofence, and one or more autonomous capabilities of the AV; and
operate the systems of the AV based in part on the impact on the error event likelihood.
2 . The system of claim 1 , wherein each error event of the one or more error events comprises at least one of an AV reroute, a failure to complete or perform an autonomous maneuver associated with a scene feature, an inability to continue the autonomous operation of the AV for at least a threshold period of time, and an inability to continue the autonomous operation of the AV without assistance from a human.
3 . The system of claim 1 , wherein determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path comprises:
for each route path of each route, determining, based on one or more scene features along the route path, an error event likelihood indicating a likelihood that the one or more scene features will cause an error event for the AV if the AV encounters the one or more scene features during the autonomous operation of the AV; and
determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path based on the error event likelihood of each route path.
4 . The system of claim 3 , wherein determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path further comprises:
determining, for each route path of each route, an estimated likelihood that the AV will use the route path for the trip comprising the route; and
determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path based on a combination of the error event likelihood of each route path and the estimated likelihood for each route path of each respective set of route paths of each route.
5 . The system of claim 3 , wherein the one or more processors are configured to:
determine the likelihood that the one or more scene features will cause the error event for the AV, the likelihood being determined based on at least one of a simulated AV trip including the one or more scene features and a previous trip of the AV along one or more routes comprising the one or more scene features.
6 . The system of claim 3 , wherein the one or more processors are configured to:
determine the likelihood that the one or more scene features will cause the error event for the AV, the likelihood being determined based on at least one of an autonomous capability of the AV, one or more maneuvers associated with the one or more scene features, and one or more operating constraints set in a software of the AV.
7 . The system of claim 6 , wherein the autonomous capability of the AV comprises at least one of a maneuver capability, a software capability, a navigation or routing capability, a sensing or perception capability, and a mechanical capability, wherein the one or more operating constraints comprise at least one of a software constraint, a routing restriction, a lane change restriction, a merge restriction, an operating restriction associated with one or more types of scenes, and a maneuver restriction.
8 . The system of claim 1 , wherein the one or more processors are configured to: determine, for each route path of each route, a number of scene features in each route
path predicted to cause an error event for the AV during the autonomous operation of the AV, wherein the error event comprises at least one of an AV reroute, a failure to complete or perform an autonomous maneuver, an inability to continue the autonomous operation of the AV for at least a threshold period of time, and an inability to continue the autonomous operation of the AV without assistance from a human;
determine a respective number of error events of each route path of each route based on the number of scene features in each route path predicted to cause the error event for the AV during the autonomous operation of the AV; and
determine an error event rate of each route path of each route based on the respective number of error events of each route path and a respective distance of each route path.
9 . The system of claim 1 , wherein the one or more processors are configured to: determine whether to implement the proposed change to the software of the AV, the sensing capability of the AV, the one or more constraints of the AV, the one or more parameters of the ODD, and the one or more autonomous capabilities of the AV based on the impact the simulation of the proposed change has on the error event likelihood of the geofence.
10 . A method for operating an autonomous vehicle (AV) comprising:
determining, via one or more processors coupled to a memory storing instructions for execution by the one or more processors, the instructions defining a simulation framework configured to generate simulations of the AV using models of its real-world operation within a model of a real-world geofenced operational design domain (ODD), wherein the one or more processors generate a plurality of simulated routes for autonomous operation of the AV located within the ODD;
determining, for each simulated route from the plurality of simulated routes, respective route paths from a starting location of the route to a destination location of the route;
determining, for each simulated route path of each route, a first respective likelihood that the AV will use the route path for an AV trip comprising the route;
determining, for each route path of each simulated route, a second respective likelihood that the AV will encounter one or more error events along the route path during the AV trip;
determining, for each route path of each simulated route, an overall error event likelihood based on the first respective likelihood and the second respective likelihood determined for each route path of each respective set of route paths of that simulated route; and
determining an aggregated error event likelihood for each simulated route based on the overall error event likelihood of each route path of that simulated route;
determining an error event likelihood for the geofence based on the aggregated error event likelihood for each simulated route; and
determining an impact on the error event likelihood of the geofence by a simulation of at least one of a proposed change to a software of the AV, a sensing capability of the AV, one or more constraints of the AV, one or more parameters of an operating design domain (ODD) associated with the geofence, and one or more autonomous capabilities of the AV; and
operating the systems of the AV based at least in part on the impact on the error event likelihood.
11 . The method of claim 10 , wherein each error event of the one or more error events comprises at least one of an AV reroute, a failure to complete or perform an autonomous maneuver associated with a scene feature, an inability to continue the autonomous operation of the AV for at least a threshold period of time, and an inability to continue the autonomous operation of the AV without assistance from a human.
12 . The method of claim 10 , wherein determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path comprises:
for each route path of each route, determining, based on one or more scene features along the route path, an error event likelihood indicating a likelihood that the one or more scene features will cause an error event for the AV if the AV encounters the one or more scene features during the autonomous operation of the AV; and
determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path based on the error event likelihood of each route path.
13 . The method of claim 12 , wherein determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path further comprises:
determining, for each route path of each route, an estimated likelihood that the AV will use the route path for the trip comprising the route; and
determining, for each route path of each route, the second respective likelihood that the AV will encounter the one or more error events along the route path based on a combination of the error event likelihood of each route path and the estimated likelihood for each route path of each respective set of route paths of each route.
14 . The method of claim 12 , further comprising:
determining the likelihood that the one or more scene features will cause the error event for the AV, the likelihood being determined based on at least one of a simulated AV trip including the one or more scene features and a previous trip of the AV along one or more routes comprising the one or more scene features.
15 . The method of claim 12 , further comprising:
determining the likelihood that the one or more scene features will cause the error event for the AV, the likelihood being determined based on at least one of an autonomous capability of the AV, one or more maneuvers associated with the one or more scene features, and one or more operating constraints set in a software of the AV, wherein the autonomous capability of the AV comprises at least one of a maneuver capability, a software capability, a navigation or routing capability, a sensing or perception capability, and a mechanical capability, wherein the one or more operating constraints comprise at least one of a software constraint, a routing restriction, a lane change restriction, a merge restriction, an operating restriction associated with one or more types of scenes, and a maneuver restriction.
16 . The method of claim 10 , further comprising:
determining, for each route path of each route, a number of scene features in each route path predicted to cause an error event for the AV during the autonomous operation of the AV, wherein the error event comprises at least one of an AV reroute, a failure to complete or perform an autonomous maneuver, an inability to continue the autonomous operation of the AV for at least a threshold period of time, and an inability to continue the autonomous operation of the AV without assistance from a human;
determining a respective number of error events of each route path of each route based on the number of scene features in each route path predicted to cause the error event for the AV during the autonomous operation of the AV; and
determining an error event rate of each route path of each route based on the respective number of error events of each route path and a respective distance of each route path.
17 . The method of claim 10 , further comprising:
determining whether to implement the proposed change to the software of the AV, the sensing capability of the AV, the one or more constraints of the AV, the one or more parameters of the ODD, and the one or more autonomous capabilities of the AV based on the impact the simulation of the proposed change has on the error event likelihood of the geofence.
18 . A non-transitory computer-readable medium comprising instructions which define a simulation framework configured to generate simulations of an autonomous vehicle (AV) using models of its real-world operation within a model of a real-world geofenced operational design domain (ODD) that, when executed by one or more processors, cause the one or more processors to:
determine a plurality of simulated routes for autonomous operation of the AV located within the ODD;
determine, for each simulated route from the plurality of routes, respective route paths from a starting location of the route to a destination location of the route;
determine, for each route path of each simulated route, a first respective likelihood that the AV will use the route path for an AV trip comprising the route;
determine, for each route path of each simulated route, a second respective likelihood that the AV will encounter one or more error events along the route path during the AV trip;
determine, for each route path of each simulated route, an overall error event likelihood based on the first respective likelihood and the second respective likelihood determined for each route path of each respective set of route paths of that simulated route; and
determine an aggregated error event likelihood for each simulated route based on the overall error event likelihood of each route path of that simulated route;
determine an error event likelihood for the geofence based on the aggregated error event likelihood for each route;
determine an impact on the error event likelihood of the geofence by a simulation of at least one of a proposed change to a software of the AV, a sensing capability of the AV, one or more constraints of the AV, one or more parameters of an operating design domain (ODD) associated with the geofence, and one or more autonomous capabilities of the AV; and
determine whether to implement the proposed change to the software of the AV, the sensing capability of the AV, the one or more constraints of the AV, the one or more parameters of the ODD, and the one or more autonomous capabilities of the AV based on the impact the simulation of the proposed change has on the error event likelihood of the geofence; and
operate the systems of the AV based on whether the proposed change has been implemented to minimize the error event likelihood.