Tracking object path in map prior layer

Abstract

Systems, methods, and devices are disclosed for predicting behaviors of objects (vehicles, bicycles, pedestrians, etc.) at a location. A model descriptive of a possible object behavior can be received by an autonomous vehicle, where the model provides conditional predictions about a future behavior of an object based on a position of the object in a lane. The autonomous vehicle can detect the position of a specific object in the lane, and the model can then be applied to determine probabilities of a future behavior of the specific object.

What is claimed is: 
     
         1 . A non-transitory computer-readable medium comprising instructions that are effective to cause at least one processor to:
 receive a map descriptive of observed vehicle paths, wherein the map provides observed paths that have been taken by multiple vehicles in a map segment, and wherein the map describes probable vehicle approach speeds from a specific lane at a specific geographic location;   as an autonomous vehicle approaches the specific lane, receive probabilities of future approach speeds of a specific vehicle from a prediction model based on an approach speed profile of the vehicle in the specific lane;   generate a planned path for the autonomous vehicle to cause the autonomous vehicle to approach the path according to parameters of the approach speed profile; and   pilot the autonomous vehicle along the planned path.   
     
     
         2 . The non-transitory computer-readable medium of  claim 1 , further comprising instructions effective to cause the at least one processor to:
 analyze sensor data descriptive of attributes of the specific vehicle, wherein at least one attribute is a signaling status, and the signaling status indicates that no turn signal is active.   
     
     
         3 . The non-transitory computer-readable medium of  claim 2 , further comprising instructions effective to cause the at least one processor to:
 determine, based on an application of the prediction model, that there is a probability, greater than a threshold, that the specific vehicle will change lanes even though the signaling status indicates that no turn signal is active.   
     
     
         4 . The non-transitory computer-readable medium of  claim 1 , wherein the map describes probable vehicle approach speeds from a specific lane at a specific geographic location 
     
     
         5 . The non-transitory computer-readable medium of  claim 1 , wherein map data is created from an aggregation of sensor data received from a fleet of vehicles, wherein the aggregation of the sensor data is used to track positions of objects over time and apply a statistical analysis to provide probabilities to probable object paths. 
     
     
         6 . The non-transitory computer-readable medium of  claim 1 , further comprising instructions effective to cause the at least one processor to:
 as an autonomous vehicle approaches an environmental feature, determine a first planned path that brings the autonomous vehicle to a stop at or prior to the environmental feature;   as the autonomous vehicle approaches the environmental feature, detect a position of the specific vehicle separate from the autonomous vehicle, wherein the specific vehicle is detected to be in a lane represented by the map segment;   apply the prediction model to determine probabilities of future paths of the specific vehicle at the specific lane, wherein at least one of the future paths has a yield probability that predicts that the specific vehicle will stop past the environmental feature; and   generate a second planned path for the autonomous vehicle by modifying the first planned path to account for the determined probabilities of the future paths of the specific vehicle, including the determined yield probability that the specific vehicle will stop past the environmental feature.   
     
     
         7 . A method comprising:
 receive a map descriptive of observed vehicle paths, wherein the map provides observed paths that have been taken by multiple vehicles in a map segment, and wherein the map describes probable vehicle approach speeds from a specific lane at a specific geographic location;   as an autonomous vehicle approaches the specific lane, receive probabilities of future approach speeds of a specific vehicle from a prediction model based on an approach speed profile of the vehicle in the specific lane;   generate a planned path for the autonomous vehicle to cause the autonomous vehicle to approach the path according to parameters of the approach speed profile; and   pilot the autonomous vehicle along the planned path.   
     
     
         8 . The method of  claim 7 , the method further comprising:
 receiving sensor data describing a specific position of the specific vehicle at a first time, and a specific position of the specific vehicle at a second time; and   aggregating the sensor data to track the specific positions of the specific vehicle at the first time and the second time.   
     
     
         9 . The method of  claim 7 , wherein the map describes probable vehicle approach speeds from a specific lane at a specific geographic location 
     
     
         10 . The method of  claim 7 , wherein mapping historical information includes an index of observed paths of vehicles from positions of the vehicles within the lane to a new lane. 
     
     
         11 . The method of  claim 7 , wherein historical information is used to create an index of observed object paths. 
     
     
         12 . The method of  claim 7 , the method further comprising:
 sending the model of object behavior to an autonomous vehicle, wherein the autonomous vehicle can use the model of object behavior to react to the specific vehicle that is proximate to the autonomous vehicle in the lane.   
     
     
         13 . The method of  claim 7 , wherein the prediction is further based on a width of a lane. 
     
     
         14 . A system comprising:
 a non-transitory computer-readable medium comprising instructions stored thereon; and   at least one processor, wherein the instructions are effective to cause the at least one processor to:
 receive a map descriptive of observed vehicle paths, wherein the map provides observed paths that have been taken by multiple vehicles in a map segment, and wherein the map describes probable vehicle approach speeds from a specific lane at a specific geographic location; 
 as an autonomous vehicle approaches the specific lane, receive probabilities of future approach speeds of a specific vehicle from a prediction model based on an approach speed profile of the vehicle in the specific lane; 
 generate a planned path for the autonomous vehicle to cause the autonomous vehicle to approach the path according to parameters of the approach speed profile; and 
 pilot the autonomous vehicle along the planned path. 
   
     
     
         15 . The system of  claim 14 , further comprising instructions effective to cause the at least one processor to:
 analyze sensor data descriptive of attributes of the specific vehicle, wherein at least one attribute is a signaling status, and the signaling status indicates that no turn signal is active.   
     
     
         16 . The system of  claim 15 , further comprising instructions effective to cause the at least one processor to:
 determine, based on an application of the prediction model, that there is a probability, greater than a threshold, that the specific vehicle will change lanes even though the signaling status indicates that no turn signal is active.   
     
     
         17 . The system of  claim 14 , wherein the map describes probable vehicle approach speeds from a specific lane at a specific geographic location 
     
     
         18 . The system of  claim 14 , wherein the map was created from an aggregation of sensor data received from a fleet of vehicles, wherein the aggregation of the sensor data is used to track positions of vehicles over time and apply a statistical analysis to provide probabilities to the probable vehicle paths. 
     
     
         19 . The system of  claim 14 , further comprising instructions effective to cause the at least one processor to:
 as an autonomous vehicle approaches an environmental feature, determine a first planned path that brings the autonomous vehicle to a stop at or prior to the environmental feature;   as the autonomous vehicle approaches the environmental feature, detect a position of the specific vehicle separate from the autonomous vehicle, wherein the specific vehicle is detected to be in a lane represented by the map segment;   apply the prediction model to determine probabilities of future paths of the specific vehicle at the specific lane, wherein at least one of the future paths has a yield probability that predicts that the specific vehicle will stop past the environmental feature; and   generate a second planned path for the autonomous vehicle by modifying the first planned path to account for the determined probabilities of the future paths of the specific vehicle, including the determined yield probability that the specific vehicle will stop past the environmental feature.   
     
     
         20 . The non-transitory computer-readable medium of  claim 19 , wherein the second planned path for the autonomous vehicle stops at an intersection beyond a location designated by a stop sign or stop line.