Abstract: The subject disclosure relates to techniques for optimizing autonomous vehicle activities at a fleet level. A process of the disclosed technology can include measuring a first set of cost metrics for a first plurality of autonomous vehicles (AVs) associated with a first vehicle routing version, wherein the first set of cost metrics comprises energy consumption values for two or more of the first plurality of AVs, measuring a second set of cost metrics for a second plurality of AVs associated with a second vehicle routing version, wherein the second set of cost metrics comprises energy consumption values for two or more of the second plurality of AVs, and comparing the first set of cost metrics with the second set of cost metrics to determine routing updates configured to reduce fleetwide energy consumption.

Abstract: The disclosed technology provides solutions for managing states of a simulated environment. In some aspects, a processor may be configured to provide a simulation environment for an autonomous vehicle (AV), wherein the simulation environment is associated with a plurality of simulation parameters each having a corresponding value dependent upon a state of the simulation environment. In some cases, the processor may determine at least one function for assessing a value of each of the plurality of simulation parameters at any time within a simulation time period. In some examples, the processor may receive a first query for a first value of at least one simulation parameter from the plurality of simulation parameters at a first time within the simulation time period. In some cases, the processor may return the first value of the at least one simulation parameter at the first time based on the at least one function.

Abstract: The disclosed technology provides solutions for improving simulation environments. In some aspects, a processor may be configured to select one or more location properties associated with a simulated road environment for testing an action of an autonomous vehicle. In some cases, the processor may identify a plurality of geographic locations having the one or more location properties associated with the simulated road environment. In some examples, the processor may generate the simulated road environment using map data associated with a first geographic location selected from the plurality of geographic locations. In some aspects, the processor may test the action of the autonomous vehicle using the simulated road environment.

Abstract: The subject disclosure relates to techniques for optimizing autonomous vehicle activities at a fleet level. A process of the disclosed technology can include measuring a first set of cost metrics for a first plurality of autonomous vehicles (AVs) associated with a first vehicle routing version, wherein the first set of cost metrics comprises energy consumption values for two or more of the first plurality of AVs, measuring a second set of cost metrics for a second plurality of AVs associated with a second vehicle routing version, wherein the second set of cost metrics comprises energy consumption values for two or more of the second plurality of AVs, and comparing the first set of cost metrics with the second set of cost metrics to determine routing updates configured to reduce fleetwide energy consumption.