Autonomous vehicle planning

A computer system comprising:

• memory having embodied thereon computer-readable instructions;
• and
• one or more hardware processors coupled to the memory,
• the computer-readable instructions configured, when executed on the one or more hardware processors, to cause the computer system to:
• plan an autonomous vehicle (AV) manoeuver in an encountered driving scenario, by executing a tree search algorithm to determine a sequence of AV manoeuvres corresponding to a path through a constructed game tree;
• wherein the constructed game tree has a plurality of nodes representing anticipated states of an encountered driving scenario, and the anticipated driving scenario state of each child node is determined by updating the driving scenario state of its parent node based on (i) a candidate AV manoeuvre and (ii) an anticipated behaviour of at least one external agent in the encountered driving scenario;
• and
• wherein the anticipated behaviour of the external agent is simulated by applying an inverse planning method to one or more observed parameters of the external agent, wherein the inverse planning method comprises:
• determining a set of available goals or manoeuvers for an external agent in the encountered driving scenario,
• for each goal or manoeuver of the set of available goals or manoeuvers, determining an expected trajectory model,
• comparing an observed trace of the external agent over a time interval with the expected trajectory model for each goal or manoeuver of the set of available goals or manoeuvers, to determine a likelihood of the goal or manoeuver, and
• using the determined likelihood of at least one of goal or manoeuver of the set of available goals or manoeuvers to simulate the anticipated behaviour of the external agent.

A method of planning autonomous vehicle (AV) manoeuvers in an encountered driving scenario, the method comprising:

• planning an AV manoeuver in an encountered driving scenario, by executing a tree search algorithm to determine a sequence of AV manoeuvres corresponding to a path through a constructed game tree;
• wherein the constructed game tree has a plurality of nodes representing anticipated states of the encountered driving scenario, and the anticipated driving scenario state of each child node is determined by updating the driving scenario state of its parent node based on (i) a candidate AV manoeuvre and (ii) an anticipated behaviour of at least one external agent in the encountered driving scenario;
• and
• wherein the anticipated behaviour of the external agent is simulated by applying a generative behaviour model to one or more observed parameters of the external agent as derived from sensor inputs, the generative behaviour model being a machine learning (ML) model which has been trained based on examples of observed, real-world driving behaviour.