Boston Dynamics, Inc.

Google Inc.

Eric Cary Whitman

Gina Christine Fay

Alex Yu Khripin

Benjamin Swilling

Alfred Anthony Rizzi

Stephen Berard

Matthew Jacob Klingensmith

Gene Brown Merewether

Christopher Stathis

Marco Da Silva

Matthew David Malchano

Adam Komoroski

Max Bajracharya

Dion Gonano

Jeffrey Yu

A computer-implemented method when executed by data processing hardware causes the data processing hardware to perform operations. The operations include detecting a candidate support surface at an elevation less than a current surface supporting a legged robot. A determination is made on whether the candidate support surface includes an area of missing terrain data within a portion of an environment surrounding the legged robot, where the area is large enough to receive a touchdown placement for a leg of the legged robot. If missing terrain data is determined, at least a portion of the area of missing terrain data is classified as a no-step region of the candidate support surface. The no-step region indicates a region where the legged robot should avoid touching down a leg of the legged robot.

Detecting negative obstacles

A method of constrained mobility mapping includes receiving from at least one sensor of a robot at least one original set of sensor data and a current set of sensor data. Here, each of the at least one original set of sensor data and the current set of sensor data corresponds to an environment about the robot. The method further includes generating a voxel map including a plurality of voxels based on the at least one original set of sensor data. The plurality of voxels includes at least one ground voxel and at least one obstacle voxel. The method also includes generating a spherical depth map based on the current set of sensor data and determining that a change has occurred to an obstacle represented by the voxel map based on a comparison between the voxel map and the spherical depth map. The method additional includes updating the voxel map to reflect the change to the obstacle.

Constrained mobility mapping

A method for terrain and constraint planning a step plan includes receiving, at data processing hardware of a robot, image data of an environment about the robot from at least one image sensor. The robot includes a body and legs. The method also includes generating, by the data processing hardware, a body-obstacle map, a ground height map, and a step-obstacle map based on the image data and generating, by the data processing hardware, a body path for movement of the body of the robot while maneuvering in the environment based on the body-obstacle map. The method also includes generating, by the data processing hardware, a step path for the legs of the robot while maneuvering in the environment based on the body path, the body-obstacle map, the ground height map, and the step-obstacle map.

Terrain aware step planning system

A method of footstep contact detection includes receiving joint dynamics data for a swing phase of a swing leg of the robot, receiving odometry data indicative of a pose of the robot, determining whether an impact on the swing leg is indicative of a touchdown of the swing leg based on the joint dynamics data and an amount of completion of the swing phase, and determining when the impact on the swing leg is not indicative of the touchdown of the swing leg, a cause of the impact based on the joint dynamics data and the odometry data.

Footstep contact detection

A computer-implemented method when executed by data processing hardware of a legged robot causes the data processing hardware to perform operations including receiving sensor data corresponding to an area including at least a portion of a docking station. The operations include determining an estimated pose for the docking station based on an initial pose of the legged robot relative to the docking station. The operations include identifying one or more docking station features from the received sensor data. The operations include matching the one or more identified docking station features to one or more known docking station features. The operations include adjusting the estimated pose for the docking station to a corrected pose for the docking station based on an orientation of the one or more identified docking station features that match the one or more known docking station features.

Robust docking of robots with imperfect sensing

A method for negotiating stairs includes receiving image data about a robot maneuvering in an environment with stairs. Here, the robot includes two or more legs. Prior to the robot traversing the stairs, for each stair, the method further includes determining a corresponding step region based on the received image data. The step region identifies a safe placement area on a corresponding stair for a distal end of a corresponding swing leg of the robot. Also prior to the robot traversing the stairs, the method includes shifting a weight distribution of the robot towards a front portion of the robot. When the robot traverses the stairs, the method further includes, for each stair, moving the distal end of the corresponding swing leg of the robot to a target step location where the target step location is within the corresponding step region of the stair.

Robotically negotiating stairs

The disclosure provides a method for generating a joint command. The method includes receiving a maneuver script including a plurality of maneuvers for a legged robot to perform where each maneuver is associated with a cost. The method further includes identifying that two or more maneuvers of the plurality of maneuvers of the maneuver script occur at the same time instance. The method also includes determining a combined maneuver for the legged robot to perform at the time instance based on the two or more maneuvers and the costs associated with the two or more maneuvers. The method additionally includes generating a joint command to control motion of the legged robot at the time instance where the joint command commands a set of joints of the legged robot. Here, the set of joints correspond to the combined maneuver.

Robot choreographer

A method for a stair tracking for modeled and perceived terrain includes receiving, at data processing hardware, sensor data about an environment of a robot. The method also includes generating, by the data processing hardware, a set of maps based on voxels corresponding to the received sensor data. The set of maps includes a ground height map and a map of movement limitations for the robot. The map of movement limitations identifies illegal regions within the environment that the robot should avoid entering. The method further includes generating a stair model for a set of stairs within the environment based on the sensor data, merging the stair model and the map of movement limitations to generate an enhanced stair map, and controlling the robot based on the enhanced stair map or the ground height map to traverse the environment.

Stair tracking for modeled and perceived terrain

A dynamic planning controller receives a maneuver for a robot and a current state of the robot and transforms the maneuver and the current state of the robot into a nonlinear optimization problem. The nonlinear optimization problem is configured to optimize an unknown force and an unknown position vector. At a first time instance, the controller linearizes the nonlinear optimization problem into a first linear optimization problem and determines a first solution to the first linear optimization problem using quadratic programming. At a second time instance, the controller linearizes the nonlinear optimization problem into a second linear optimization problem based on the first solution at the first time instance and determines a second solution to the second linear optimization problem based on the first solution using the quadratic programming. The controller also generates a joint command to control motion of the robot during the maneuver based on the second solution.

Dynamic planning controller

A legged robot may seek to operate according to a target gait. The legged robot may include leg members and leg joints. Possibly based on the target gait and state of the legged robot, an ordered list of gait controllers may be obtained. The gait controllers in the ordered list may define respective gaits of the legged robot, and may include respective validity checks and output parameters for the respective gaits. The ordered list may begin with a target gait controller that defines the target gait. The ordered list may be traversed in order from the target gait controller until a validity check associated with a particular gait controller passes. The legged robot may be instructed to actuate the leg members and/or leg joints according to output parameters of the particular gait controller.

Achieving a target gait behavior in a legged robot

A method of planning a swing trajectory for a leg of a robot includes receiving an initial position of a leg of the robot, an initial velocity of the leg, a touchdown location, and a touchdown target time. The method also includes determining a difference between the initial position and the touchdown location and separating the difference between the initial position and the touchdown location into a horizontal motion component and a vertical motion component. The method also includes selecting a horizontal motion policy and a vertical motion policy to satisfy the motion components. Each policy produces a respective trajectory as a function of the initial position, the initial velocity, the touchdown location, and the touchdown target time. The method also includes executing the selected policies to swing the leg of the robot from the initial position to the touchdown location at the touchdown target time.

Leg swing trajectories

An example implementation involves controlling robots with non-constant body pitch and height. The implementation involves obtaining a model of the robot that represents the robot as a first point mass rigidly coupled with a second point mass along a longitudinal axis. The implementation also involves determining a state of a first pair of legs, and determining a height of the first point mass based on the model and the state of the first pair of legs. The implementation further involves determining a first amount of vertical force for at least one leg of the first pair of legs to apply along a vertical axis against a surface while the at least one leg is in contact with the surface. Additionally, the implementation involves causing the at least one leg of the first pair of legs to begin applying the amount of vertical force against the surface.

Control of robotic devices with non-constant body pitch

A method for perception and fitting for a stair tracker includes receiving sensor data for a robot adjacent to a staircase. For each stair of the staircase, the method includes detecting, at a first time step, an edge of a respective stair of the staircase based on the sensor data. The method also includes determining whether the detected edge is a most likely step edge candidate by comparing the detected edge from the first time step to an alternative detected edge at a second time step, the second time step occurring after the first time step. When the detected edge is the most likely step edge candidate, the method includes defining, by the data processing hardware, a height of the respective stair based on sensor data height about the detected edge. The method also includes generating a staircase model including stairs with respective edges at the respective defined heights.

Growing community of inventors

Watertown, MA, United States of America

Eric Cary Whitman