Traffic control system

A traffic control system for use at a junction, the traffic control system comprising:

• a plurality of sensors for monitoring vehicles and/or other road users at and around the junction;
• a traffic control agent subsystem;
• and
• traffic signals including signal outputs for controlling the vehicles and/or other users of the junction,
• the sensors providing inputs to the traffic control agent subsystem, and the traffic control agent subsystem controlling the traffic signals to optimise traffic flow in accordance with one or more goals,
• in which the traffic control agent subsystem includes:
• a junction simulation model of the junction being controlled, the junction simulation model accepting inputs of traffic scenarios and accepting inputs of control decisions, and providing outputs of traffic patterns as a result of the control decisions made;
• an agent training system, in which traffic control machine learning agents are trained on the junction simulation model by reinforcement learning to optimise traffic flow in accordance with one or more goals, in which a batch of master agents is created for training in the agent training subsystem, each member of the batch of agents having a control strategy which is at least in part randomly defined, in which multiple copies of the batch of master agents are created as multiple batches of worker agents, each batch of worker agents taking part in a training workout, each training workout including a different scenario or different scenarios, and the multiple workouts taking place in parallel, at the same time as each other;
• and
• a live traffic control system, the live traffic control system including a traffic control machine learning agent which has been trained in the agent training system, the traffic control machine learning agent in the live traffic control system accepting inputs from the sensors and sending outputs to the traffic signals to control the traffic at the junction.