Patents by Inventor Arpan Kusari

Arpan Kusari has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).

  • CERTIFIED ADVERSARIAL ROBUSTNESS FOR DEEP REINFORCEMENT LEARNING
    Publication number: 20210103800
    Abstract: The present disclosure describes systems and methods that include calculating one or more lower bound state-action values based on a corrupted observation and a predetermined perturbation parameter; and selecting an action corresponding to a lower bound state-action value having the highest value.
    Type: Application
    Filed: October 7, 2019
    Publication date: April 8, 2021
    Applicants: Ford Global Technologies, LLC, MIT Technology Licensing Office
    Inventors: Bjoern Malte Luetjens, Michael F. Everett, Jonathan P. How, Arpan Kusari
  • Distinguishing virtual objects from one another
    Patent number: 10916014
    Abstract: A system includes a processor and memory storing instructions executable by the processor. The instructions include determining whether a currently-evaluated point is part of a segment. This may include: determining feature vectors for the currently-evaluated point and for each of a plurality of previously-evaluated points tagged to the segment; using the feature vectors to determine a probability score; and determine whether the score is greater than a threshold.
    Type: Grant
    Filed: June 1, 2018
    Date of Patent: February 9, 2021
    Assignee: FORD GLOBAL TECHNOLOGIES, LLC
    Inventor: Arpan Kusari
  • Vehicle neural network processing
    Patent number: 10831208
    Abstract: A computing system can be programmed to determine a vehicle action based on vehicle sensor data input to a deep neural network (DNN) trained using an inverse reinforcement learning (IRL) system that includes a variational auto-encoder (VAE). The computing system can be further programmed to operate a vehicle based on the vehicle action.
    Type: Grant
    Filed: November 1, 2018
    Date of Patent: November 10, 2020
    Assignee: FORD GLOBAL TECHNOLOGIES, LLC
    Inventor: Arpan Kusari
  • Vehicle adaptive learning
    Patent number: 10733510
    Abstract: A computing system can determine a vehicle action based on inputting vehicle sensor data to a first neural network including a first safety agent that can determine a probability of unsafe vehicle operation. The first neural network can be adapted, at a plurality of times, by a periodically retrained deep reinforcement learning agent that includes a second deep neural network including a second safety agent. A vehicle can be operated based on the vehicle action.
    Type: Grant
    Filed: August 24, 2018
    Date of Patent: August 4, 2020
    Assignee: FORD GLOBAL TECHNOLOGIES, LLC
    Inventors: Subramanya Nageshrao, Hongtei Eric Tseng, Dimitar Petrov Filev, Ryan Lee Baker, Christopher Cruise, Leda Daehler, Shankar Mohan, Arpan Kusari
  • Vehicle action control
    Patent number: 10703370
    Abstract: One or more target areas are identified proximate to a moving vehicle. The vehicle can be maneuvered to a target area selected according to a reinforcement learning reward function.
    Type: Grant
    Filed: August 24, 2018
    Date of Patent: July 7, 2020
    Assignee: FORD GLOBAL TECHNOLOGIES, LLC
    Inventors: Arpan Kusari, Hongtei Eric Tseng
  • VEHICLE LANGUAGE PROCESSING
    Publication number: 20200142420
    Abstract: A computing system can be programmed to determine a vehicle action based on vehicle sensor data input to a deep neural network (DNN) trained using an inverse reinforcement learning (IRL) system that includes a variational auto-encoder (VAE). The computing system can be further programmed to operate a vehicle based on the vehicle action.
    Type: Application
    Filed: November 1, 2018
    Publication date: May 7, 2020
    Applicant: Ford Global Technologies, LLC
    Inventor: Arpan Kusari
  • VEHICLE ADAPTIVE LEARNING
    Publication number: 20200065665
    Abstract: A computing system can determine a vehicle action based on inputting vehicle sensor data to a first neural network including a first safety agent that can determine a probability of unsafe vehicle operation. The first neural network can be adapted, at a plurality of times, by a periodically retrained deep reinforcement learning agent that includes a second deep neural network including a second safety agent. A vehicle can be operated based on the vehicle action.
    Type: Application
    Filed: August 24, 2018
    Publication date: February 27, 2020
    Applicant: Ford Global Technologies, LLC
    Inventors: Subramanya Nageshrao, Hongtei Eric Tseng, Dimitar Petrov Filev, Ryan Lee Baker, Christopher Cruise, Leda Daehler, Shankar Mohan, Arpan Kusari
  • VEHICLE ACTION CONTROL
    Publication number: 20200062262
    Abstract: One or more target areas are identified proximate to a moving vehicle. The vehicle can be maneuvered to a target area selected according to a reinforcement learning reward function.
    Type: Application
    Filed: August 24, 2018
    Publication date: February 27, 2020
    Applicant: Ford Global Technologies, LLC
    Inventors: Arpan Kusari, Hongtei Eric Tseng
  • DISTINGUISHING VIRTUAL OBJECTS FROM ONE ANOTHER
    Publication number: 20190370975
    Abstract: A system includes a processor and memory storing instructions executable by the processor. The instructions include determining whether a currently-evaluated point is part of a segment. This may include: determining feature vectors for the currently-evaluated point and for each of a plurality of previously-evaluated points tagged to the segment; using the feature vectors to determine a probability score; and determine whether the score is greater than a threshold.
    Type: Application
    Filed: June 1, 2018
    Publication date: December 5, 2019
    Applicant: Ford Global Technologies, LLC
    Inventor: ARPAN KUSARI
  • DEPTH-DATA SEGMENTATION
    Publication number: 20190197167
    Abstract: An imaging system is described, as well as methods of using the system. A method may include: parsing a grid-tree data structure for a plurality of previously-evaluated elevation points; determining that none of the plurality are within a radius of a currently-evaluated elevation point; and then, generating a new segment comprising the currently-evaluated elevation point.
    Type: Application
    Filed: December 21, 2017
    Publication date: June 27, 2019
    Applicant: Ford Global Technologies, LLC
    Inventors: Arpan Kusari, Thomas Ackenhausen