System for predicting aggressive driving

A system for predicting aggressive driving behavior for a driver of a vehicle, the system comprising:

• a sensor array that is a part of a traffic infrastructure and located stationary near a road intersection, the sensor array is configured to output spatial-temporal data indicative an object and movement of the object around or about the traffic intersection, wherein the object includes the vehicle;
• one or more edge computing devices configured to acquire short-term time-series data from at least one of (i) the sensor array and (ii) one or more vehicle sensors in the vehicle, each of the one or more edge computing devices comprising:
• a distributed cloud networking system having one or more trained models, wherein the one or more trained models includes one or more machine learning models, and wherein the one more trained models include at least a first trained model to predict and describe a trajectory of the vehicle;
• a control element;
• and
• a roadside unit is in communication with the distributed cloud networking system and the control element and is integrated into the traffic infrastructure and located stationary near or about a road intersection, the roadside unit having:
• a processor, and
• a non-transitory computer-readable medium including instructions that are executable by the processor of the one or more edge computing devices, the instructions included on the non-transitory computer-readable medium comprising:
• predicting an approach speed of the vehicle to an intersection based on the short-term time-series data,
• generating a driving pattern corresponding to driving style based on the short-term time-series data wherein the driving style is based on measures of vehicle motion, an energy consumption, and a force on the vehicle,
• identifying aggressive driving behavior based on the driving patterns,
• predicting, using one or more artificial intelligence algorithms, a predicted trajectory of the vehicle based on the predicted approach speed of the vehicle,
• predicting a real-time aggressive driving behavior based on the driving pattern, the identified aggressive driving behavior, and the predicted trajectory of the vehicle;
• and
• wherein the instructions further comprise controlling a control element in real-time to mitigate the predicted real-time aggressive driving behavior, in response to predicting the real-time aggressive driving behavior, and wherein the control element includes at least one of controlling one or more traffic control signals to the traffic infrastructure,
• each of the one or more edge computing devices comprises:
• a speed prediction model is a machine learning model and is configured to receive the short-term time-series data and long-term time-series data, wherein the short-term times-series data is a different type of data than the long-term time-series data, wherein the short-term time-series data is configured to change at least one of:
• every time stamp, time interval or a combination thereof, and wherein the long-term time-series data changes over a longer period of time than the short-term time-series data, and wherein the speed prediction model includes:
• a plurality of encoders includes at least a first encoder and a second encoder, wherein each of the plurality of encoders includes a plurality of recurrent units, wherein:
• a first encoder of the plurality of the encoders is configured to acquire a multivariate short-term time-series data periodically at a short-term interval, and propagating the multivariate short-term time-series data through a first plurality of recurrent units of the first encoder to output a first hidden state parameter from the first encoder, and
• a second encoder of the plurality of encoders configured to acquire a univariate short-term time-series data periodically at the short-term interval and propagate the univariate short-term time-series data through a second plurality of recurrent units of the second encoder to output a second hidden state parameter from the second encoder,
• an artificial neural network configured to acquire a multivariate long-term time-series data periodically at a long-term interval, and
• a decoder having a stack of recurrent units, wherein each of the recurrent units includes a machine learning module configured to predict vehicle speed at a future time step and each of a predicted vehicle speed is associated with a different future time.

A system for predicting aggressive driving for a driver of a vehicle, the system comprising:

• one or more edge computing devices configured to acquire short-term time-series data and long-term time-series data from at least one of (i) one or more sensors that are part of traffic infrastructure and (ii) one or more vehicle sensors in the vehicle, each of the one or more edge computing devices having:
• a speed prediction model is configured to receive at the short-term time-series data and the long-term time-series data, wherein:
• the speed prediction model is a machine learning model,
• the short-term times-series data is a different type of data than the long-term time-series data, and
• the short-term time-series data is configured to change at least one of:
• every time stamp, time interval or a combination thereof,
• wherein the speed prediction model includes:
• a plurality of encoders includes at least a first encoder and a second encoder, the first encoder includes a first plurality of recurrent units and the second encoder includes a second plurality of recurrent units, said first encoder different than the second encoder, and
• a decoder;
• and
• wherein a speed prediction model further includes:
• a processor, and
• a non-transitory computer-readable medium including instructions that are executable by the processor of the one or more edge computing devices, the instructions included on the non-transitory computer-readable medium comprising:
• acquiring a multivariate short-term time-series data periodically at a short-term interval,
• acquiring a univariate short-term time-series data periodically at the short-term interval,
• acquiring a multivariate long-term time-series data periodically at a long-term interval,
• propagating the multivariate short-term time-series data through a first plurality of recurrent units of the first encoder to output a first hidden state parameter from the first encoder,
• propagating the univariate short-term time-series data through a second plurality of recurrent units of the second encoder to output a second hidden state parameter from the second encoder;
• processing the multivariate long-term time-series data with a neural network,
• concatenating the first hidden state parameter, the hidden state parameter, and the processed multivariate long-term time-series data to generate an intermediate vector,
• inputting the intermediate vector to the decoder, and
• predicting a plurality of vehicle speeds from the decoder based on the intermediate vector.