Toyota Research Institute, Inc.

Brandon Northcutt

Konstantine Mushegian

Katarina Bouma

Kevin Stone

Christopher Gregg

Astrid Glende Jackson

Akiyoshi Ochiai

Ewaldo Eder Carvalho Santana, Jr

Kazuto Murase

Astrid Jackson

Yasuyuki Kohaya

Stephen Hughes

A method for training an object detection system includes estimating a location of a first object in an environment based on a density cluster map generated from a plurality of images of the environment. The method also includes generating one or more negative training samples of the first object in the environment based on the plurality of images, each of the one or more negative training samples corresponding to a second object at a location in the environment that is different than the estimated location of the first object. The method further includes generating positive training samples from a set of images of the first object. The method also includes training the object detection system to detect the first object based on the positive training samples and the negative training sample.

Object detection improvement based on autonomously selected training samples

A vehicle includes an electronic control unit that includes a processor; and one or more processor-readable instructions; and an occupant detection sensor configured to output a signal indicative of a presence of an occupant. When executed by the processor, the one or more processor-readable instructions cause the vehicle to: record a route from a first location to a second location, and autonomously return to the first location from the second location using the recorded route based on the signal received from the occupant detection sensor.

Autonomous rideshare rebalancing

A method for controlling a robotic device includes observing a first object associated with an object type at one or more first locations in an environment over a period of time prior to a current time. The method also includes generating a probability distribution associated with the one or more first locations based on observing the first object over the period of time. The method further includes observing, at the current time, a second object associated with the object type at a second location in the environment. The method still further includes determining a probability of the second object being at the second location based on observing the second object at the second location. The probability is based on the probability distribution associated with the one or more first locations. The method also includes controlling the robotic device to perform an action based on the probability being less than a threshold.

Fall detection and assistance

A method for controlling a robotic device based on observed object locations includes obtaining a group of observations of an object type in an environment based on identifying one or more first objects associated with the object type over a period of time in the environment. The method also includes localizing the object type in the environment based on a location cluster comprising a group of nodes. Each node in the group of nodes associated with one observation of the group of observations. The method further includes generating a cost map indicating a probability distribution of the object type in relation to the localized object type in the environment. The method still further includes controlling the robotic device to perform an action in the environment based on the cost map and a map of the environment.

Object location analysis

A method for generating positive and negative training samples is presented. The method includes identifying false positive images of an object based on multiple images of an environment. The method also includes generating positive training samples from a set of images of the object. The method further includes generating a negative training sample from the false positive image. The method still further includes training an object detection system based on the positive training samples and the negative training sample.

A method for controlling a robotic device based on observed object locations is presented. The method includes observing objects in an environment. The method also includes generating a probability distribution for locations of the observed objects. The method further includes controlling the robotic device to perform an action in the environment based on the generated probability distribution.

A system for determining when a vehicle has learned a route is provided. The system includes at least one camera configured to capture image data of an external environment, a processor and a memory module storing one or more processor-readable instructions that, when executed by the processor, cause the system to: generate one or more predictions about objects in the external environment as the vehicle proceeds along the route based on the image data, determine a prediction error based on an accuracy of the one or more predictions, determine a confidence level for the route based on the prediction error, and determine that the vehicle has learned the route in response to the confidence level exceeding a confidence level threshold.

Systems and methods for training a vehicle to autonomously drive a route

A method for controlling a robotic device based on observed object locations is presented. The method includes observing objects in an environment. The method also includes generating a probability distribution for locations of the observed objects. The method further includes controlling the robotic device to perform an action when an object is at a location in the environment with a location probability that is less than a threshold.

Systems, robots, and methods for generating three-dimensional skeleton representations of people are disclosed. A method includes generating, from a two-dimensional image, a two-dimensional skeleton representation of a person present in the two-dimensional image. The two-dimensional skeleton representation includes a plurality of joints and a plurality of links between individual joints of the plurality of joints. The method further includes positioning a cone around one or more links of the plurality of links, and identifying points of a depth cloud that intersect with the cone positioned around the one or more links of the two-dimensional skeleton. The points of the depth cloud are generated by a depth sensor and each point provides depth information. The method also includes projecting the two-dimensional skeleton representation into three-dimensional space using the depth information of the points that intersect with the cone, thereby generating the three-dimensional skeleton representation of the person.

Growing community of inventors

Arlington, MA, United States of America

Brandon Northcutt