Autonomous vehicle simulation and code build scheduling

Abstract

Systems and methods for autonomous vehicle (AV) simulation and code build scheduling are provided. A method includes receiving a first task specification for a first task associated with a first AV simulation and/or a first AV code build, receiving, a second task specification for a second task associated with a second AV simulation and/or a second AV code build, and executing a portion of the first task concurrently with a portion of the second task based on the portion of the first task and the portion of the second task have different resource requirements. The portion of the first task is associated with one of an AV asset download, an AV code execution, or an AV artifact upload. The portion of the second task is associated with a different one of the AV asset download, the AV code execution, or the AV artifact upload.

What is claimed is: 
     
         1 . A method comprising:
 receiving, by a computer-implemented system, a first job request to execute one or more tasks associated with at least one of a first autonomous vehicle (AV) simulation or a first AV code build, the first job request including:
 a task specification specifying AV driving scenario data for a first task of the one or more tasks; and 
 a job completion deadline associated with the one or more tasks; 
   scheduling, by the computer-implemented system using a completion time-driven model, one or more workers to execute the first task within the job completion deadline; and   transmitting, by the computer-implemented system to the one or more workers, the task specification.   
     
     
         2 . The method of  claim 1 , wherein the completion time-driven model includes a task runtime model trained on task runtime data associated with at least one of a second AV simulation or a second AV code build different from the at least one of the first AV simulation or the first AV code build. 
     
     
         3 . The method of  claim 2 , further comprising:
 receiving, by the computer-implemented system, a completion indication for the first task; and   updating, by the computer-implemented system, the task runtime model based on a task runtime and the task specification for the first task, wherein the task runtime is based at least in part on the completion indication.   
     
     
         4 . The method of  claim 2 , wherein the scheduling comprises:
 generating, using the task runtime model, an estimated runtime for the first task based on the task specification; and   calculating an estimated job completion time based at least in part on the estimated runtime.   
     
     
         5 . The method of  claim 4 , wherein the generating the estimated runtime for the first task is further based on the task specification including at least one of:
 information for downloading AV assets including at least the AV driving scenario data;   information for downloading an executable image associated with the AV driving scenario data; or   information for uploading AV artifacts.   
     
     
         6 . The method of  claim 1 , further comprising:
 receiving, by the computer-implemented system, a second job request to execute at least a second task associated with at least one of a second AV simulation or a second AV code build, the second job request including:
 a task specification for the second task; and 
 a job completion deadline associated with the second task, 
   wherein the scheduling comprises:
 scheduling, using the completion time-driven model, a first worker of the one or more workers to execute the second task before the first task based on the job completion deadline associated with the second task being earlier than the job completion deadline associated with the first task. 
   
     
     
         7 . The method of  claim 1 , wherein:
 a first worker of the one or more workers has a capability higher than a resource requirement specified by the task specification for the first task; and   the scheduling comprises:
 scheduling, based on an availability of the first worker, the first worker to execute the first task. 
   
     
     
         8 . The method of  claim 1 , wherein the scheduling comprises:
 determining, based at least in part on a runtime for the first task and a remaining time to the job completion deadline, whether to schedule a preemptible worker or a non-preemptible worker of the one or more workers to execute the first task.   
     
     
         9 . The method of  claim 8 , wherein the determining whether to schedule the preemptible worker or the non-preemptible worker to execute the first task is further based on at least one of:
 a threshold number of allowable preemptible workers;   a threshold number of committed non-preemptible workers;   a threshold number of allowable non-preemptible workers;   a number of queued tasks including the first task; or   a number of workers to execute the queued tasks.   
     
     
         10 . The method of  claim 9 , wherein the scheduling the one or more workers further comprises:
 scheduling, in response to the determining, the preemptible worker to execute the first task; and   scheduling, in response to a failure to complete the first task on the preemptible worker, the non-preemptible worker to execute the first task.   
     
     
         11 . One or more non-transitory, computer-readable media encoded with instructions that, when executed by one or more processing units, perform a method comprising:
 receiving a first job request to execute one or more tasks associated with at least one of a first autonomous vehicle (AV) simulation or a first AV code build, the first job request including:
 a task specification specifying AV driving scenario data for a first task of the one or more tasks; and 
 a job completion deadline associated with the one or more tasks; 
   scheduling, using a completion time-driven model, one or more workers to execute the first task within the job completion deadline; and   transmitting, to the one or more workers, the task specification.   
     
     
         12 . The one or more non-transitory, computer-readable media of  claim 11 , wherein the completion time-driven model includes a task runtime model trained on task runtime data associated with at least one of a second AV simulation or a second AV code build different from the at least one of the first AV simulation or the first AV code build. 
     
     
         13 . The one or more non-transitory, computer-readable media of  claim 12 , the method further comprising:
 receiving a completion indication for the first task; and   updating the task runtime model based on a task runtime and the task specification for the first task, wherein the task runtime is based at least in part on the completion indication.   
     
     
         14 . The one or more non-transitory, computer-readable media of  claim 12 , wherein the scheduling comprises:
 generating, using the task runtime model, an estimated runtime for the first task based on the task specification; and   calculating an estimated job completion time based at least in part on the estimated runtime.   
     
     
         15 . The one or more non-transitory, computer-readable media of  claim 14 , wherein the generating the estimated runtime for the first task is further based on the task specification including at least one of:
 information for downloading AV assets including at least AV driving scenario data;   information for downloading an executable image associated with the AV driving scenario data; or   information for uploading AV artifacts.   
     
     
         16 . A computer-implemented system comprising:
 one or more processing units; and   one or more non-transitory computer-readable media storing instructions, when executed by the one or more processing units, cause the one or more processing units to perform operations comprising:
 receiving a first job request to execute one or more tasks associated with at least one of a first autonomous vehicle (AV) simulation or a first AV code build, the first job request including:
 a task specification specifying AV driving scenario data for a first task of the one or more tasks; and 
 a job completion deadline associated with the one or more tasks; scheduling, using a completion time-driven model, one or more workers to 
 
   execute the first task within the job completion deadline; and
 transmitting, to the one or more workers, the task specification. 
   
     
     
         17 . The computer-implemented system of  claim 16 , wherein the completion time-driven model includes a task runtime model trained on task runtime data associated with at least one of a second AV simulation or a second AV code build different from the at least one of the first AV simulation or the first AV code build. 
     
     
         18 . The computer-implemented system of  claim 17 , the operations further comprising: 
 receiving a completion indication for the first task; and   updating the task runtime model based on a task runtime and the task specification for the first task, wherein the task runtime is based at least in part on the completion indication.   
     
     
         19 . The computer-implemented system of  claim 17 , wherein the scheduling comprises:
 generating, using the task runtime model, an estimated runtime for the first task based on the task specification; and   calculating an estimated job completion time based at least in part on the estimated runtime.   
     
     
         20 . The computer-implemented system of  claim 19 , wherein the generating the estimated runtime for the first task is further based on the task specification including at least one of:
 information for downloading AV assets including at least AV driving scenario data;   information for downloading an executable image associated with the AV driving scenario data; or   information for uploading AV artifacts.