Abstract: A method includes of routing an autonomous vehicle includes receiving information obtained from a camera on a second vehicle distinct from the autonomous vehicle. The method includes automatically identifying a road condition using image analysis of the information received from the camera on the second vehicle. The method includes receiving a request to route the autonomous vehicle from a first location to a second location; and in response to the request: generating a cost model for routing the autonomous vehicle, wherein the cost model includes a cost of the road condition automatically identified from the information received from the camera on the second vehicle; selecting a route from the first location to the second location in accordance with the cost model; and routing an autonomous vehicle in accordance with the selected route.

Abstract: A method includes of routing an autonomous vehicle includes receiving information obtained from a camera on a second vehicle distinct from the autonomous vehicle. The method includes automatically identifying a road condition using image analysis of the information received from the camera on the second vehicle. The method includes receiving a request to route the autonomous vehicle from a first location to a second location; and in response to the request: generating a cost model for routing the autonomous vehicle, wherein the cost model includes a cost of the road condition automatically identified from the information received from the camera on the second vehicle; selecting a route from the first location to the second location in accordance with the cost model; and routing an autonomous vehicle in accordance with the selected route.

Abstract: A method includes storing representations of a passengers and fleet vehicles. The method includes generating a graph representation of a geographic map that includes requested pick-up locations and drop-off locations for the passengers and generating a state graph representation of the passengers and the fleet vehicles. The state graph representation includes a plurality of nodes connected by edges. Each of the plurality of nodes represents a candidate state of the passengers and the fleet vehicles. A respective edge of the state graph representation represents an action of a respective vehicle picking up or dropping off a passenger. The respective edge has a cost that is based at least in part on traversal of the graph representation of the geographic map. The method further includes using the state graph representation to generate a set of routes and route the fleet vehicles in accordance with the generated set of routes.

Abstract: A method includes generating a cost model for routing an autonomous vehicle. The cost model includes, for an intersection, a plurality of costs for distinct maneuvers through the intersection. The cost model is represented as a graph of nodes and edges, the edges having respective edge weights that represent costs in the cost model. Each edge of the representation of the intersection represents a respective maneuver through the intersection. The method further includes, in response to a request to route an autonomous vehicle: selecting a route in accordance with the cost model, wherein the selecting is based at least in part on one of the plurality of costs for the distinct maneuvers through the intersection; and routing the autonomous vehicle in accordance with the selected route.