Apparatus and method for assisting an autonomous vehicle and/or a driver of a vehicle

1 . An apparatus for assisting a control system and/or a driver of a vehicle, the apparatus comprising:

processing circuitry configured to:

determine, by implementing an artificial neural network and based on motion state information of the vehicle, a first heat map image representing probabilities that the vehicle will be at respective positions of a plurality of positions at one or more future time instants, and

determine, by implementing the artificial neural network and based on motion state information of at least one traffic object, a second heat map image representing probabilities that the at least one traffic object will be at the respective positions of the plurality of positions at the one or more future time instants;

determine for the plurality of positions, based on the first heat map image and the second heat map image, a third heat map image representing probabilities that the vehicle and the at least one traffic object will be at the respective positions of the plurality of positions at the one or more future time instants; and

control, based on the third heat map image indicating a risk of collision that exceeds a threshold, the vehicle to:

cause a deceleration and/or a change in a route of the vehicle, and/or

issue a visual or acoustical warning signal to the driver of the vehicle,

wherein the artificial neural network comprises a social encoder configured to:

receive encoded past motion state information of the vehicle and encoded past motion state information of the at least one traffic object, and

generate, using an attention mechanism, encoded interactions between the vehicle and the at least one traffic object.

2 . The apparatus of claim 1 , wherein in determining the respective first heat map, the artificial neural network is configured to decode a feature vector, generated based on the encoded interactions between the vehicle and the at least on traffic object, by a multi-head decoder to:

determine a plurality of potential routes of the vehicle;

determine a respective likelihood for each of the plurality of potential routes of the vehicle; and

determine, for the plurality of positions and based on the plurality of potential routes of the vehicle and/or the respective likelihood for each of the plurality of potential routes of the vehicle, the respective first heat map image.

3 . The apparatus of claim 1 , wherein the artificial neural network further comprises a map encoder configured to process map information about the vicinity of the vehicle, and

wherein the neural network is configured to decode a feature vector, generated based on both (i) the encoded interactions between the vehicle and the at least one traffic object and (ii) the encoded map information, by a multi-head decoder to:

determine, for the plurality of positions, the respective first heat map image.

4 . The apparatus of claim 3 , wherein the map information comprises a map image, and wherein the map encoder comprises a convolutional neural network configured to process the map image.

5 . The apparatus of claim 1 , wherein the neural network further comprises a first history encoder configured to process past motion state information of the vehicle to generate the encoded past motion state information of the vehicle and a second history encoder configured to process past motion state information of the at least one traffic object to generate the encoded past motion state information of the at least one traffic object.

6 . The apparatus of claim 1 , wherein the visual or acoustical warning signal is configured to assist the driver of the vehicle to identify the at least one traffic object within the vicinity of the vehicle.

7 . The apparatus of claim 1 , wherein the processing circuitry is further configured to detect the at least one traffic object in a vicinity of the vehicle based on one or more images of the vicinity of the vehicle.

8 . The apparatus of claim 1 , wherein the motion state information of the vehicle comprises at least one of a position, a velocity, an acceleration, a motion direction and a route of the vehicle.

9 . The apparatus of claim 1 , wherein the motion state information of the at least one traffic object comprises at least one of a position, a velocity, an acceleration, a motion direction and a route of the at least one traffic object.

10 . A vehicle comprising an apparatus according to claim 1 .

11 . The vehicle of claim 10 , wherein the vehicle comprises a head-up display (HUD) or a console screen configured to display the visual or acoustical warning signal to the driver of the vehicle.

12 . The vehicle of claim 10 , wherein the vehicle further comprises a dashboard camera configured to capture one or more images of the vicinity of the vehicle for detecting the at least one traffic object in the vicinity of the vehicle.

13 . A cloud server comprising the apparatus of claim 1 , wherein the cloud server is configured to communicate with the vehicle via a wireless communication network.

14 . The cloud server of claim 13 , wherein the cloud server is configured to receive, from a plurality of further vehicles, further motion state information of the plurality of further vehicles and/or further motion state information of a plurality of further traffic objects within a respective vicinity of the plurality of further vehicles, and to determine a further plurality of combined probability measures based on the further motion state information.

15 . A method for assisting a control system and/or a driver of a vehicle, the method comprising:

determine, by implementing an artificial neural network and based on motion state information of the vehicle, a first heat map image representing probabilities that the vehicle will be at respective positions of a plurality of positions at one or more future time instants, and

determine, by implementing the artificial neural network and based on motion state information of at least one traffic object, a second heat map image representing probabilities that the at least one traffic object will be at the respective positions of the plurality of positions at the one or more future time instants;

determining for the plurality of positions, based on the first heat map image and the second heat map image, a third heat map image representing probabilities that the vehicle and the at least one traffic object will be at the respective positions of the plurality of positions at the one or more future time instants; and

controlling, based on the third heat map image indicating a risk of collision that exceeds a threshold, the vehicle to:

cause a deceleration and/or a change in a route of the vehicle, and/or

issue a visual or acoustical warning signal to the driver of the vehicle,

wherein the artificial neural network comprises a social encoder configured to:

receive encoded past motion state information of the vehicle and encoded past motion state information of the at least one traffic object, and

generate, using an attention mechanism, encoded interactions between the vehicle and the at least one traffic object.

16 . A non-transitory computer-readable storage medium having stored thereon processor-executable code that, upon execution by a computer or a processor, causes the computer or the processor to perform the method of claim 15 .