Soongsil University

Ewha University-Industry Collaboration Foundation

Minhae Kwon

Dongsu Lee

Hyunggon Park

Nayoung Kim

Hyoseon Kye

The present disclosure relates to a driving characteristic inference apparatus and method of a vehicle. According to an exemplary embodiment of the present disclosure, a driving characteristic inference method includes observing a target driving vehicle which is an observation target to generate driving data including behavior information of the target driving vehicle, by an observation module; performing reinforcement learning on an artificial intelligence model using learning data including a first driving characteristic coefficient, by a learning module; generating sampled inference behavior data with the driving data and the first driving character coefficient as an input of the artificial intelligence model, by a model utilization module; and comparing the generated inference behavior data with measured actual behavior data to determine the second driving characteristic coefficient, by an inference module.

Apparatus and method for inferring driving characteristics of a vehicle

A method of building an ad-hoc network of a wireless relay node and an ad-hoc network system are disclosed. The method includes verifying state information representing a relative distance and angle to a neighboring node capable of receiving data when each of relay nodes transmits the data, determining an action representing a change over time of each relay node, determining, based on an amount of change in a network throughput determined according to the state information and an amount of energy consumption according to the action, a reward corresponding to the action, and building a network including a source node, a destination node, and a plurality of relay nodes by generating, based on a reward of each of the relay nodes, a policy that allows a cumulative reward to be maximized.

Method of building ad-hoc network of wireless relay node and ad-hoc network system

A method for combating a stop-and-go wave problem using deep reinforcement learning based autonomous vehicles includes selecting one of a plurality of deep reinforcement learning algorithms for training an autonomous vehicle and a reward function in a roundabout environment in which autonomous vehicles and non-autonomous vehicles are driving, determining a deep neural network architecture according to the selected deep reinforcement learning algorithm, learning a policy which enables the autonomous vehicle to drive at a closest velocity to a constant velocity based on state information including a velocity of the autonomous vehicle and a relative velocity and a relative position between the autonomous vehicle and an observable vehicle by the autonomous vehicle at a preset time interval and reward information, using the selected deep reinforcement learning algorithm, and driving the autonomous vehicle based on the learned policy to determine an action of the autonomous vehicle.

Method for combating stop-and-go wave problem using deep reinforcement learning based autonomous vehicles, recording medium and device for performing the method

An anomaly detection method includes searching for one principal component axis by analyzing a normal data set collected in time series from a plurality of IoT devices by using a principal component analysis technique, setting a center point of the principal component, receiving a currently measured measurement data set from the plurality of IoT devices, acquiring a linear transformation data set having a plurality of projection points as elements by projecting a plurality of measurement data which is each element in the measurement data set onto the principal component axis, calculating a Mahalanobis distance between the projection point and the central point, and detecting whether or not data of the IoT devices is abnormal by comparing the Mahalanobis distance calculated for each element with a threshold.

Growing community of inventors

Seoul, South Korea

Minhae Kwon