Path planning method using vehicle risk model trained on manual intervention driving data

1 . A path planning method applied to an autonomous vehicle, the autonomous vehicle configured with a vehicle risk model, the path planning method comprising:

obtaining historical driving data from the autonomous vehicle and historical environmental perception data, wherein the historical driving data comprises historical actual driving data and historical planned driving data generated by the vehicle risk model;

setting data differences between the historical actual driving data and the historical planned driving data as manual intervention driving data;

determining first perception data matching with the manual intervention driving data in the historical environmental perception data, and determining risk level information matching with the manual intervention driving data according to the first perception data;

detecting whether the risk level information indicates an emergency risk or a non-emergency risk;

in response to the risk level information indicating the emergency risk, training the vehicle risk model based on the risk level information, the first perception data, the historical planned driving data, and the manual intervention driving data;

in response to the risk level information indicating the non-emergency risk, assigning a first priority to the manual intervention driving data when the non-emergency risk is a medium risk, or assigning a second priority to the manual intervention driving data when the non-emergency risk is a low risk, wherein a level of the first priority is higher than a level of the second priority; sorting the manual intervention driving data based on the assigned priority, and adding the manual intervention driving data to a training task list; and training the vehicle risk model according to the training task list, wherein the manual intervention driving data having the first priority is extracted from the training task list for training the vehicle risk model every first preset interval, and the manual intervention driving data having the second priority is extracted from the training task list for training the vehicle risk model every second preset interval, the second preset interval is longer than the first preset interval; and

generating a planned path to drive the autonomous vehicle based on the trained vehicle risk model in response to receiving a vehicle travel planning request.

2 . The path planning method of claim 1 , wherein determining the risk level information according to the first perception data comprises:

determining the emergency risk in response to the first perception data indicating that, without manual intervention, a collision would occur; and

determining the non-emergency risk in response to the first perception data indicating that, without manual intervention, the autonomous vehicle would experience a non-collision driving anomaly.

3 . The path planning method of claim 1 , wherein the manual intervention driving data comprises at least one of: a rotation angle and a rotation direction of a steering wheel under manual intervention, a tread distance of a foot brake under manual intervention, and a trampling trip of an accelerator under manual intervention.

4 . The path planning method of claim 1 , wherein the first perception data comprises at least one of:

dynamic environmental perception data and static environmental perception data, wherein the dynamic environmental perception data comprises movement information of dynamic obstacles and traffic signal information, the static environmental perception data comprises a width of a road and location information of static obstacles.