May Mobility, Inc.

University of Michigan

Toyota Research Institute, Inc.

Ford Global Technolgoies, LLC

Edwin B Olson

Tom Voorheis

Collin Johnson

Steve Vozar

Sajan Patel

Jeff Sterniak

Melinda Kothbauer

Sean M Messenger

Ryan M Eustice

Gregory Meyer

Stuart Fairley

Asheesh Ranjan

Dhanvin H Mehta

Gonzalo Ferrer

Rob Goeddel

Michael R James

Andrew P Richardson

James McBride

Akshay Katpatal

Kapil Borle

Yash Bagla

Enric Galceran

Mahmood Mahmood

Robert Goeddel

Ryan W Wolcott

Alexander Glencairn Cunningham

Johannes Strom

John Wang

John Mamish

Carl Kershaw

Jacob Crossman

John H Marcoux

Timothy Saucer

Alexander Cunningham

Ahmed Elshaarany

Mitesh Agrawal

Alan Heirich

A system for impact-based operation of an autonomous agent (equivalently referred to herein as an ego agent and autonomous vehicle) includes and/or interfaces with a computing subsystem (equivalently referred to herein as a computer and/or set of computers). A method for impact-based operation of an autonomous agent includes: receiving a set of inputs; predicting a set of future scenarios; and determining a set of metrics based on the set of future scenarios. Additionally or alternatively, the method can include operating the autonomous agent based on the set of metrics and/or any other processes.

Method and system for impact-based operation of an autonomous agent

A system and method includes an autonomous agent having a communication interface that enables the autonomous agent to communicate with a plurality of infrastructure sensing devices; a plurality of distinct health monitors that monitor distinct operational aspects of the autonomous agent; an autonomous state machine that computes a plurality of allowed operating states of the autonomous agent based on inputs from the plurality of distinct health monitors; a plurality of distinct autonomous controllers that generate a plurality of distinct autonomous control instructions; and an arbiter of autonomous control instructions that: collects, as a first input, the plurality of autonomous control instructions generated by each of the plurality of distinct autonomous controllers; collects, as a second input, data relating to the plurality of allowed operating state of the autonomous agent; and selectively enables only a subset of the autonomous control instructions to pass to driving components of the autonomous agent.

Systems and methods for implementing multimodal safety operations with an autonomous agent

The method for dynamically updating an environmental representation of an autonomous agent can include: receiving a set of inputs S; generating an environmental representation S; and updating the environmental representation S. Additionally or alternatively, the method Scan include providing the environmental representation to a planning module Sand/or any other suitable processes. The method Sfunctions to generate and/or dynamically update an environmental representation to facilitate control of an autonomous agent.

Method and system for dynamically updating an environmental representation of an autonomous agent

A system for operating an autonomous agent with incomplete environmental information can include and/or interface an autonomous operating system and an autonomous agent. A method for operating an autonomous agent with incomplete environmental information includes any or all of: receiving a set of inputs; determining a set of known objects in the ego vehicle's environment; determining a set of blind regions in the ego vehicle's environment; and inserting a set of virtual objects into the set of blind regions; selecting a set of virtual objects based on the set of blind regions; operating the autonomous agent based on the set of virtual objects; and/or any other suitable processes.

Method and system for operating an autonomous agent with incomplete environmental information

In Multi-Policy Decision-Making (MPDM), many computationally-expensive forward simulations are performed in order to predict the performance of a set of candidate policies. In risk-aware formulations of MPDM, only the worst outcomes affect the decision making process, and efficiently finding these influential outcomes becomes the core challenge. Recently, stochastic gradient optimization algorithms, using a heuristic function, were shown to be significantly superior to random sampling. In this disclosure, it was shown that accurate gradients can be computed-even through a complex forward simulation—using approaches similar to those in dep networks. The proposed approach finds influential outcomes more reliably, and is faster than earlier methods, allowing one to evaluate more policies while simultaneously eliminating the need to design an easily-differentiable heuristic function.

Method and apparatus for constructing informative outcomes to guide multi-policy decision making

The method can include: receiving a set of inputs; determining a set of policies based on the set of inputs; determining a set of scores associated with the set of environmental policies; and evaluating the set of policies. Additionally or alternatively, the method can include operating the ego agent according to a selected policy and/or any other processes. The method functions to facilitate scoring of policies based on 'feasibility' for agents in an environment. Additionally or alternatively, the method can function to facilitate autonomous operation of a vehicle (e.g., based on policy-feasibility of agents in the environment). Additionally or alternatively, the method can function to facilitate intention estimation for agents in an environment.

Method and system for feasibility-based operation of an autonomous agent

A systemfor remote assistance of an autonomous agent can include and/or interface with any or all of: a sensor suite, a computing system, a communication interface, and/or any other suitable components. The system can further optionally include a set of infrastructure devices, a teleoperator platform, and/or any other suitable components. The systemfunctions to enable information to be exchanged between an autonomous agent and a tele-assist. Additionally or alternatively, the systemcan function to operate the autonomous agent (e.g., based on remote inputs received from a teleoperator, indirectly, etc.) and/or can perform any other suitable functions.

Method and system for remote assistance of an autonomous agent

A system for intelligently implementing an autonomous agent that includes an autonomous agent, a plurality of infrastructure devices, and a communication interface. A method for intelligently calibrating infrastructure (sensing) devices using onboard sensors of an autonomous agent includes identifying a state of calibration of an infrastructure device, collecting observation data from one or more data sources, identifying or selecting mutually optimal observation data, specifically localizing a subject autonomous agent based on granular mutually optimal observation data, identifying dissonance in observation data from a perspective of a subject infrastructure device, and recalibrating a subject infrastructure device.

Systems and methods for intelligently calibrating infrastructure devices using onboard sensors of an autonomous agent

A system and a method for autonomous decisioning and operation by an autonomous agent includes: collecting decisioning data including: collecting a first stream of data includes observation data obtained by onboard sensors of the autonomous agent, wherein each of the onboard sensors is physically arranged on the autonomous agent; collecting a second stream of data includes observation data obtained by offboard infrastructure devices, the offboard infrastructure devices being arranged geographically remote from and in an operating environment of the autonomous agent; implementing a decisioning data buffer that includes the first stream of data from the onboard sensors and the second stream of data from the offboard sensors; generating current state data; generating/estimating intent data for each of one or more agents within the operating environment of the autonomous agent; identifying a plurality of candidate behavioral policies; and selecting and executing at least one of the plurality of candidate behavioral policies.

Multi-perspective system and method for behavioral policy selection by an autonomous agent

A methodassessing and mitigating risks encounterable by an autonomous vehicle includes collecting information associated with an environment of an ego vehicle and determining and assessing a set of risks encounterable by the ego vehicle. A system for assessing and mitigating risks can include and/or interface with an ego vehicle (equivalently referred to herein as an autonomous vehicle, autonomous agent, ego agent, agent, etc.) and a set of computing subsystems (equivalently referred to herein as a set of computers) and/or processing subsystems (equivalently referred to herein as a set of processors), which function to implement any or all of the processes of the method.

Method and system for assessing and mitigating risks encounterable by an autonomous vehicle

A system for dynamic policy curation includes a computing system and interfaces with an autonomous agent. A method for dynamic policy curation includes collecting a set of inputs; processing the set of inputs; and determining a set of available policies based on processing the set of inputs. Additionally or alternatively, the method can include any or all of: selecting a policy; implementing a policy; and/or any other suitable processes.

Method and system for dynamically curating autonomous vehicle policies

In Multi-Policy Decision-Making (MPDM), many computationally-expensive forward simulations are performed in order to predict the performance of a set of candidate policies. In risk-aware formulations of MPDM, only the worst outcomes affect the decision making process, and efficiently finding these influential outcomes becomes the core challenge. Recently, stochastic gradient optimization algorithms, using a heuristic function, were shown to be significantly superior to random sampling. In this disclosure, it was shown that accurate gradients can be computed—even through a complex forward simulation—using approaches similar to those in dep networks. The proposed approach finds influential outcomes more reliably, and is faster than earlier methods, allowing one to evaluate more policies while simultaneously eliminating the need to design an easily-differentiable heuristic function.

Growing community of inventors

Ann Arbor, MI, United States of America

Edwin B Olson