Decision making for motion control

Abstract

A behavior planner for a vehicle generates a plurality of conditional action sequences of the vehicle using a tree search algorithm and heuristics obtained from one or more machine learning models. Each sequence corresponds to a sequence of anticipated states of the vehicle. At least some of the action sequences are provided to a motion selector of the vehicle. The motion selector generates motion-control directives based on the received conditional action sequences and on data received from one or more sensors of the vehicle, and transmits the directives to control subsystems of the vehicle.

1 .- 20 . (canceled) 
     
     
         21 . A method, comprising:
 generating, at a behavior planner comprising one or more computing devices, a plurality of conditional action sequences of a vehicle, wherein individual ones of the conditional action sequences correspond to a respective set of anticipated states of the vehicle, and wherein the plurality of conditional action sequences includes a first conditional action sequence and a second conditional action sequence;   evaluating, by the behavior planner, the first conditional action sequence with respect to the second conditional action sequence using at least respective resource consumption estimates of the first and second conditional action sequences; and   transmitting, by the behavior planner, the first conditional action sequence to a motion selector of the vehicle based at least in part on a result of the evaluating.   
     
     
         22 . The method as recited in  claim 21 , wherein evaluating the first conditional action sequence with respect to the second conditional action sequence comprises:
 comparing the respective resource consumption estimates of the first and second conditional action sequences.   
     
     
         23 . The method as recited in  claim 21 , wherein the respective resource consumption estimates comprise respective fuel consumption estimates for the first and second conditional action sequences. 
     
     
         24 . The method as recited in  claim 21 , wherein the respective resource consumption estimates comprise respective battery consumption estimates for the first and second conditional action sequences. 
     
     
         25 . The method as recited in  claim 21 , wherein evaluating the first conditional action sequence with respect to the second conditional action sequence is further based on respective environmental impact estimates of the first and second conditional action sequences. 
     
     
         26 . The method as recited in  claim 25 , further comprising:
 weighting the respective resource consumption estimates and the respective environmental impact estimates according to occupant preferences for an occupant of the vehicle.   
     
     
         27 . The method as recited in  claim 25 , wherein the respective environmental impact estimates comprise respective carbon footprint estimates of the first and second conditional action sequences. 
     
     
         28 . A system, comprising:
 a behavior planner implemented at one or more computing devices, wherein the behavior planner is configured to:
 generate a plurality of conditional action sequences of a vehicle, wherein individual ones of the conditional action sequences correspond to a respective set of anticipated states of the vehicle, and wherein the plurality of conditional action sequences includes a first conditional action sequence and a second conditional action sequence; 
 evaluate the first conditional action sequence with respect to the second conditional action sequence using at least respective resource consumption estimates of the first and second conditional action sequences; and 
 transmit the first conditional action sequence to a motion selector of the vehicle based at least in part on a result of the evaluating. 
   
     
     
         29 . The system as recited in  claim 28 , wherein to evaluate the first conditional action sequence with respect to the second conditional action sequence, the behavior planner is further configured to:
 compare the respective resource consumption estimates of the first and second conditional action sequences.   
     
     
         30 . The system as recited in  claim 28 , wherein the respective resource consumption estimates comprise respective fuel consumption estimates for the first and second conditional action sequences. 
     
     
         31 . The system as recited in  claim 28 , wherein the respective resource consumption estimates comprise respective battery consumption estimates for the first and second conditional action sequences. 
     
     
         32 . The system as recited in  claim 28 , wherein the behavior planner is configured to evaluate the first conditional action sequence with respect to the second conditional action sequence further based on respective environmental impact estimates of the first and second conditional action sequences. 
     
     
         33 . The system as recited in  claim 32 , wherein the behavior planner is configured to:
 weight the respective resource consumption estimates and the respective environmental impact estimates according to occupant preferences for an occupant of the vehicle.   
     
     
         34 . The system as recited in  claim 32 , wherein the respective environmental impact estimates comprise respective carbon footprint estimates of the first and second conditional action sequences. 
     
     
         35 . One or more non-transitory computer-accessible storage media storing program instructions that when executed on or across one or more processors implements a behavior planner for a vehicle, wherein the behavior planner is configured to:
 generate a plurality of conditional action sequences of a vehicle, wherein individual ones of the conditional action sequences correspond to a respective set of anticipated states of the vehicle, and wherein the plurality of conditional action sequences includes a first conditional action sequence and a second conditional action sequence;   evaluate the first conditional action sequence with respect to the second conditional action sequence using at least respective resource consumption estimates of the first and second conditional action sequences; and   transmit the first conditional action sequence to a motion selector of the vehicle based at least in part on a result of the evaluating.   
     
     
         36 . The one or more non-transitory computer-accessible storage media as recited in  claim 35 , wherein to evaluate the first conditional action sequence with respect to the second conditional action sequence, the behavior planner is further configured to:
 compare the respective resource consumption estimates of the first and second conditional action sequences.   
     
     
         37 . The one or more non-transitory computer-accessible storage media as recited in  claim 35 , wherein the respective resource consumption estimates comprise respective fuel consumption estimates for the first and second conditional action sequences. 
     
     
         38 . The one or more non-transitory computer-accessible storage media as recited in  claim 35 , wherein the respective resource consumption estimates comprise respective battery consumption estimates for the first and second conditional action sequences. 
     
     
         39 . The one or more non-transitory computer-accessible storage media as recited in  claim 35 , wherein the behavior planner is configured to evaluate the first conditional action sequence with respect to the second conditional action sequence further based on respective environmental impact estimates of the first and second conditional action sequences. 
     
     
         40 . The one or more non-transitory computer-accessible storage media as recited in  claim 39 , wherein the respective environmental impact estimates comprise respective carbon footprint estimates of the first and second conditional action sequences.