Systems, apparatus, and methods for embedded encodings of contextual information using a neural network with vector space modeling

An apparatus for implementing a neural network system for training software embedded in an autonomous vehicle (AV) and for creating a built-in neural network for controlling the AV, the apparatus comprising:

• at least one processor deployed on a computing device within the AV, the at least one processor being programmed to implement applications to generate control data for an AV based on sensor data from sensors deployed on the AV wherein the at least one processor having an input coupled to the sensors to receive the sensor data;
• the at least one processor having an associated memory for storing data of context to vector (context2vec) words of AV words for use in controlling the AV;
• the at least one processor having a neural network having a plurality of nodes with context to vector (context2vec) contextual embeddings to enable operations of the AV;
• the at least one processor encoding a plurality of context2vec AV words in a sequence of timing to embed data of context and behavior derived from sensor data of the AV in time sequences wherein the context and behavior data comprise:
• at least mapping, situational, and behavior data of the AV operation;
• the at least one processor with a set of inputs to the neural network which comprise:
• at least one of a current, a prior, and a subsequent encoded context2vec AV word each represented in a one-hot scheme of a set of possibilities of context2vec AV words wherein at least one context2vec AV word of the set is designated with an ON state while other context2vec AV words of the set are designated in an OFF state;
• the at least one processor executing a neural network to determine a target context2vec AV word of each set of the inputs based on the current context2vec AV word;
• the at least one processor enabling an output vector computed by the neural network that represents the embedded distributional one-hot scheme of the input encoded context2vec AV word;
• and
• the at least one processor controlling, by a set of control actions, a behavior of the AV based on the context2vec AV word.

A method for implementing a neural network system for training software embedded in an autonomous vehicle (AV) and for creating a built-in neural network for controlling the AV, said method comprising:

• encoding, by at least one computer disposed in the AV, a plurality of context2vec AV words encoded in a sequence of timing to embed data of context and behavior derived from sensor data of the AV in time sequences wherein the context and behavior data comprise:
• at least mapping, situational, and behavior data of the AV operation;
• selecting, by the at least one computer, a set of inputs to the neural network which comprise:
• at least one of a current, a prior, and a subsequent encoded context2vec AV word each represented in a one-hot scheme of a set of possibilities of context2vec AV words wherein at least one context2vec AV word of the set is designated with an ON state while other context2vec AV words of the set are designated in an OFF state;
• applying a neural network solution, by the at least one computer, to determine a target context2vec AV word of each set of the inputs based on the current context2vec AV word;
• computing an output vector, by the at least one computer, by the neural network that represents the embedded distributional one-hot scheme of the input encoded context2vec AV word;
• and
• controlling, by the at least one computer, a set of behavior control operations for controlling, a behavior of the AV based on the context2vec AV word.