Apparatuses, systems, and methods for vehicle trajectory and collision prediction
Disclosed apparatuses and methods for trajectory and collision prediction comprising a processor operable to generate, using a machine-learning (ML) model, a ML predicted trajectory of a vehicle and a ML model confidence of the ML predicted trajectory, generate, using a heuristics model, a heuristic predicted trajectory of the vehicle, determine whether the ML model confidence satisfies a metric, in response to determining that the ML model confidence satisfies the metric, apply the ML predicted trajectory as a predicted trajectory, and in response to determining that the ML learning model confidence fails the metric, apply the heuristic predicted trajectory as the predicted trajectory.
1 . An apparatus for trajectory and collision prediction comprising a processor operable to:
generate, using a machine-learning (ML) model, a ML predicted trajectory of a vehicle and a ML model confidence of the ML predicted trajectory;
generate, using a heuristics model, a heuristic predicted trajectory of the vehicle;
determine whether the ML model confidence satisfies a metric;
in response to determining that the ML model confidence satisfies the metric, apply the ML predicted trajectory as a predicted trajectory;
in response to determining that the ML learning model confidence fails the metric, apply the heuristic predicted trajectory as the predicted trajectory;
generate collision confidence associated with the predicted trajectory, wherein the collision confidence comprises a blending collision confidence generated by blending a ML collision confidence and a heuristic collision confidence using weights based on the ML model confidence;
determine whether the collision confidence is beyond a collision threshold, and
in response to determining that the collision confidence is beyond the collision threshold, generate a trajectory planning and operate the vehicle based on the trajectory planning.
2 . The apparatus of claim 1 , wherein the processor is operable to further:
in response to determining that the collision confidence is beyond the collision threshold, inform the vehicle of a predicted collision.
3 . The apparatus of claim 2 , wherein the processor is operable to further instruct the vehicle to follow the trajectory planning.
4 . The apparatus of claim 1 , wherein the ML collision confidence is generated based on the ML model in response to determining that the ML model confidence satisfies the metric.
5 . The apparatus of claim 1 , wherein the heuristic collision confidence is generated based on the heuristic model in response to determining that the ML model confidence fails the metric.
6 . The apparatus of claim 1 , wherein the blending collision confidence is determined based on following equation:
blending collision confidence=ML model confidence×ML collision confidence+(1−ML model confidence)×heuristic collision confidence.
7 . The apparatus of claim 1 , wherein the ML model confidence is determined based on a probabilistic model or a distance-based model.
8 . The apparatus of claim 1 , wherein the ML predicted trajectory and the heuristic predicted trajectory are generated based on vehicle kinematics, vehicle controls, and road conditions.
9 . The apparatus of claim 8 , wherein the vehicle kinematics comprise a vehicle position, a vehicle velocity, and a vehicle acceleration.
10 . The apparatus of claim 8 , wherein the vehicle controls comprise vehicle steering, vehicle throttle, and brake inputs.
11 . The apparatus of claim 8 , wherein the road conditions comprise road surface conditions, road geometry, and traffic conditions.
12 . The apparatus of claim 8 , wherein the apparatus further comprises one or more vision sensors and speed sensors configured to collect information of the vehicle kinematics, the vehicle controls, and the road conditions.
13 . A method for trajectory and collision prediction comprising:
generating, using a machine-learning (ML) model, a ML predicted trajectory of a vehicle and a ML model confidence of the ML predicted trajectory;
generating, using a heuristics model, a heuristic predicted trajectory of the vehicle;
determining whether the ML model confidence satisfies a metric;
in response to determining that the ML model confidence satisfies the metric, applying the ML predicted trajectory as a predicted trajectory;
in response to determining that the ML learning model confidence fails the metric, applying the heuristic predicted trajectory as the predicted trajectory;
generating collision confidence associated with the predicted trajectory, wherein the collision confidence comprises a blending collision confidence generated by blending a ML collision confidence and a heuristic collision confidence using weights based on the ML model confidence;
determining whether the collision confidence is beyond a collision threshold, and
in response to determining that the collision confidence is beyond the collision threshold, generating a trajectory planning and operating the vehicle based on the trajectory planning.
14 . The method of claim 13 , wherein the method further comprises:
in response to determining that the collision confidence is beyond the collision threshold, informing the vehicle of a predicted collision, and instruct the vehicle to follow the trajectory planning.
15 . The method of claim 13 , wherein:
the ML collision confidence is generated based on the ML model in response to determining that the ML model confidence satisfies the metric; and
the heuristic collision confidence is generated based on the heuristic model in response to determining that the ML model confidence fails the metric.
16 . The method of claim 13 , wherein the blending collision confidence is determined based on following equation:
blending collision confidence=ML model confidence×ML collision confidence+(1−ML model confidence)×heuristic collision confidence.
17 . The method of claim 13 , wherein the ML model confidence is determined based on a probabilistic model or a distance-based model.
18 . The method of claim 13 , wherein the ML predicted trajectory and the heuristic predicted trajectory are generated based on vehicle kinematics, vehicle controls, and road conditions, wherein:
the vehicle kinematics comprise a vehicle position, a vehicle velocity, and a vehicle acceleration;
the vehicle controls comprise vehicle steering, vehicle throttle, and brake inputs;
the road conditions comprise road surface conditions, road geometry, and traffic conditions; and
information of the vehicle kinematics, the vehicle controls, and the road conditions are collected by one or more vision sensors and speed sensors.