Hrl Laboratories, LLC

The Boeing Company

Michael A Warren

Christopher Serrano

Aleksey Nogin

Pape Sylla

Byron N Heersink

Deepak Khosla

Yang Chen

Heiko Hoffmann

Kevin R Martin

Gavin D Holland

Sean Soleyman

Pape Maguette Sylla

Joshua G Fadaie

Charles Richard Tullock

Ignacio M Soriano

Shawn Moffit

Byron Nicholas Heersink

Calvin Chung

Systems and methods provide for the evaluation of flight controls for an aerial vehicle and define a vehicle dynamics model corresponding to an aerial vehicle. One or more barrier functions are constructed based on the defined vehicle dynamics model, and candidate invariant sets are identified for the defined vehicle dynamics model using the one or more barrier functions. A plurality of flight controls of the aerial vehicle are analyzed using the identified candidate invariant sets and within the analyzed candidate invariant sets, command tracking of the aerial vehicle is confirmed with control commands, using the analyzed plurality of flight controls. Control commands falling outside of the analyzed candidate invariant sets are identified.

Evaluation of air vehicle flight controls

Described is an attack system for generating perturbations of input signals in a recurrent neural network (RNN) based target system using a deep reinforcement learning agent to generate the perturbations. The attack system trains a reinforcement learning agent to determine a magnitude of a perturbation with which to attack the RNN based target system. A perturbed input sensor signal having the determined magnitude is generated and presented to the RNN based target system such that the RNN based target system produces an altered output in response to the perturbed input sensor signal. The system identifies a failure mode of the RNN based target system using the altered output.

Deep reinforcement learning based method for surreptitiously generating signals to fool a recurrent neural network

Described is a system for training a neural network for estimating surface normals for use in operating an autonomous platform. The system uses a parallelizable k-nearest neighbor sorting algorithm to provide a patch of points, sampled from the point cloud data, as input to the neural network model. The points are transformed from Euclidean coordinates in a Euclidean space to spherical coordinates. A polar angle of a surface normal of the point cloud data is estimated in the spherical coordinates. The trained neural network model is utilized on the autonomous platform, and the estimate of the polar angle of the surface normal is used to guide operation of the autonomous platform within the environment.

Neural network architecture for small LIDAR processing networks for slope estimation and ground plane segmentation

Described is a system for controlling a mobile platform. A neural network that runs on the mobile platform is trained based on a current state of the mobile platform. A Satisfiability Modulo Theories (SMT) solver capable of reasoning over non-linear activation functions is periodically queried to obtain examples of states satisfying specified constraints of the mobile platform. The neural network is then trained on the examples of states. Following training on the examples of states, the neural network selects an action to be performed by the mobile platform in its environment. Finally, the system causes the mobile platform to perform the selected action in its environment.

Solving based introspection to augment the training of reinforcement learning agents for control and planning on robots and autonomous vehicles

Described is a system and method for generating safety conditions for a cyber-physical system with state space S, action space A and trajectory data labelled as either safe or unsafe. In operation, the system receives inputs and ten minimizes loss functions to cause a neural network to become a barrier function. Based on the barrier function, the system can then determine if the cyber-physical system is entering an usafe state, such that if the cyber-physical system is entering the usafe state, then the cyber-physical system is caused to initiate a maneuver to position the cyber-physical system into a safe state.

Automated system for generating approximate safety conditions for monitoring and verification

Apparatus and methods for training a machine learning algorithm (MLA) to control a first aircraft in an environment that comprises the first aircraft and a second aircraft are described. Training of the MLA can include: the MLA determining a first-aircraft action for the first aircraft to take within the environment; sending the first-aircraft action from the MLA; after sending the first-aircraft action, receiving an observation of the environment and a reward signal at the MLA, the observation including information about the environment after the first aircraft has taken the first-aircraft action and the second aircraft has taken a second-aircraft action, the reward signal indicating a score of performance of the first-aircraft action based on dynamic and kinematic properties of the second aircraft; and updating the MLA based on the observation of the environment and the reward signal.

Autonomous behavior generation for aircraft

Described is a system for automatically generating images that satisfy specific image properties. Using a code parser component, a tensor expression intermediate representation (IR) of a deep neural network code is produced. A specification describing a set of image properties is parsed in a fixed formal syntax. The tensor expression IR and the specification is input into a rewriting and analysis engine. The rewriting and analysis engine queries an external solver to obtain pixel values satisfying the specification. When pixel values satisfying the specification can be found in a fixed time period, the rewriting and analysis engine combines the pixel values into an image that satisfies the specification and outputs the image.

Automatic generation of images satisfying specified neural network classifier properties

Described is a control command system for generating control commands for a vehicle. The system receives a networked control system corresponding to a network of subsystems of sensors and actuators of the vehicle and data collected from sensors on the vehicle. Multiple subsystems are formed from the networked control system. A dynamic mode decomposition with control (DMDc) process is applied to each subsystem, yielding a linear approximation for each subsystem. The linear approximations are combined into a single linear approximation for the networked control system, and a linear control system approximating the networked control system is output. Control commands are generated for the vehicle based on the linear control system, which cause the vehicle to perform a vehicle operation.

Process to control a networked system of time varying devices

Described is a high-assurance network gateway generator that generates and encodes network gateway code on a computer readable medium. In operation, the network gateway generator receives input artifacts, which are translated into corresponding formats as translated data. The translated data is distributed to an OS code generator, a glue code generator, and a communications code generator. The OS code generator then generates OS code based on the translated data. The communications code generator proceeds to generate deserialization and filtering code based on the translated data. Further, a glue code generator generates glue code based on the OS code and translated data. An executable network gateway code is then generated by combining the glue code, deserialization code, and filtering code. Finally, the executable network gateway code is then encoded on a computer readable medium.

Growing community of inventors

Northridge, CA, United States of America

Michael A Warren