Methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting occupancies of agents. One of the methods includes obtaining scene data characterizing a current scene in an environment; and processing a neural network input comprising the scene data using a neural network to generate a neural network output, wherein: the neural network output comprises respective occupancy outputs corresponding to a plurality of agent types at one or more future time points; the occupancy output for each agent type at a first future time point comprises respective occupancy probabilities for a plurality of locations in the environment; and in the occupancy output for each agent type at the first future time point, the respective occupancy probability for each location characterizes a likelihood that an agent of the agent type will occupy the location at the first future time point.

A device may receive emergency data, traffic data, network performance data, crime data, and gunshot data associated with a geographical area and may identify a location within the geographical area based on the emergency data, the traffic data, the network performance data, the crime data, and the gunshot data. The device may determine, based on the emergency data, the traffic data, the network performance data, the crime data, and the gunshot data for the location, a risk level for the location and may identify an autonomous vehicle based on the risk level, the traffic data, and the network performance data for the location. The device may determine a route for the autonomous vehicle to the location based on the traffic data and the network performance data for the location, and may perform actions based on the autonomous vehicle and the route.

Techniques pertaining to an autonomous police vehicle are described. A method may involve a processor associated with an autonomous vehicle obtaining an indication of violation of one or more traffic laws by a first vehicle. The method may also involve the processor maneuvering the autonomous vehicle to pursue the first vehicle. The method may further involve the processor remotely executing one or more actions with respect to the first vehicle.

An autonomous vehicle (AV) includes a conveyor system arranged within its main cabin. The conveyor system includes attachment points for attaching transport bags to the conveyor system, and a motor to move the transport bags to different positions along the conveyor system. The AV includes a door that can open and close to provide access to the transport bags. The AV further includes a conveyor system controller that identifies a transport bag associated with a user and issues an instruction to the motor to move the transport bags along the conveyor system.