Scenario cluster in self-driving motor vehicle operation
A protocol and a system for customizing driving operation of and legalizing self-driving motor vehicles are introduced as an alternative solution to state-of-the-art technologies in development and marketing of the self-driving motor vehicles, promising to enable a self-driving motor vehicle to provide an experience as if being driven by the mind of a passenger the first time it serves on a public roadway.
1. A method of clustering data of a plurality of scenarios in operation of a self-driving motor vehicle comprising the steps of:
acquiring data of a respective one of the plurality of scenarios;
having the data of the respective one of the scenarios measured and quantized into a vector; respective to an operation behavior effective in the respective one of the scenarios, finding a set of data pairs comprising two boundary coordinate values of each component of the vector of the data of the respective one of the scenarios, wherein one of the two boundary coordinates values is smaller than a respective one coordinate of the vector, and the other of the two boundary coordinates values is larger than the respective one coordinate of the vector;
dividing each of the data pairs into a plurality of segments;
constructing a candidate matching scenario of the respective one of the scenarios comprised of n coordinates, wherein each of the n coordinates is a boundary value of one of the plurality of segments of a respective one of the components;
determining the candidate matching scenario being a member of a cluster of the respective one of the scenarios respective to the operation behavior in the respective one of the scenarios if the operation behavior is effective in the candidate matching scenario, or a probability for the operation behavior to be effective in the candidate matching scenario is larger than a threshold value;
wherein the self-driving motor vehicle operates according to the operation behavior in the respective one of the scenarios or in another one of the scenarios of the cluster associated with the candidate matching scenario.