Amazon Technologies, Inc.

Jong Jin Park

Daimian Wang

Rajasimman Madhivanan

Yue Hu

Roopesh Athipatla Pattabhi

Changsheng Shen

Jingyu Qiao

James Ballantyne

Joseph W Durham

Chang Young Kim

James Charles Zamiska

Mohit Deshpande

Samer Nashed

Nathalie Majcherczyk

Nicholas Chuanji Liao

Michael Cheng

Dhruva Kumar

Harry S Cheng

Babak Kianmajd

Maps of a physical space provide information to an autonomous mobile device (AMD) about the location of objects. For example, a local map may represent objects currently detected while a global map represents objects that persist over time. Disagreements between the maps may result from quantization errors, sensor errors, changes in the physical space over time that are shown in one map but not another, and so forth. For example, the local map may indicate an obstacle in a space that the global map indicates is empty, or vice versa. A cluster representing an object in a map may be assessed to distinguish if the object is actual or false. Other techniques may be used to determine disagreements. Information about persistent disagreements may be stored for mitigation. The AMD may mitigate disagreements by exploring the physical space associated with the disagreement to add to map data.

System to determine map data for autonomous mobile device map

An autonomous mobile device (AMD) has one or more joints that move. A 'choreo' comprises instructions that specify a particular set of motions. A constrained optimization algorithm determines a first time at which operation of the joint may be transitioned from a current state, such as while performing a first choreo, to a second choreo. The first time is determined to minimize motion discontinuities to less than constraint values. Constraint values may specify limits on time, position, acceleration, velocity and so forth. During the transition the joint is moved, subject to the constraints and using an optimization algorithm, toward the positions specified by the second choreo, subject to the constraint values. Synchronization of the second choreo to a specified time is maintained. The resulting movement avoids discontinuities producing a more naturalistic motion and may reduce wear and tear on the joint.

System for motion control of one or more joints

An autonomous mobile device (AMD) may move around a physical space while performing tasks. The AMD may have sensors with a field of view (FOV) facing forward. As the AMD moves forward, a safe region is determined based on data from those forward-facing sensors. The safe region describes a geographical area that is clear of obstacles during recent travel. If the AMD moves outside of the FOV, such as moving backwards, a check is made as to whether the AMD remains within the safe region. If a portion of the AMD moves beyond the safe region, the AMD stops. This prevents the AMD from colliding with obstacles that are outside of the safe region and also outside of the FOV while moving.

System for movement outside of sensor field of view by an autonomous mobile device

An autonomous mobile device (AMD) may move around a physical space to perform tasks. The AMD may explore the physical space to determine map data characterizing the physical space. An exploration system maintains a world snapshot that is representative of the AMD with respect to the physical space. A plurality of planning algorithm modules are available, each able to achieve a particular exploration goal or operate under particular conditions such as initial motion, loop closure, frontier exploration, and so forth. One or more planning algorithms are selected and provided with the world snapshot data. These planning algorithms provide responses and may also determine plan data. The resulting plan data is then selected for execution. The execution of the selected plan data operates the AMD to explore the physical space.

System for planning exploration by autonomous mobile device

An autonomous mobile device (AMD) moves around a physical space while performing tasks. The AMD may have sensors with fields of view (FOVs) that are forward-facing. As the AMD moves forward, a safe region is determined based on data from those forward-facing sensors. The safe region describes a geographical area clear of obstacles during recent travel. Before moving outside of the current FOV, the AMD determines whether a move outside of the current FOV keeps the AMD within the safe region. For example, if a path that is outside the current FOV would result in the AMD moving outside the safe region, the AMD modifies the path until poses associated with the path result in the AMD staying within the safe region. The resulting safe path may then be used by the AMD to safely move outside the current FOV.

System for path planning in areas outside of sensor field of view by an autonomous mobile device

Embodiments herein describe a robotic system that uses range sensors to identify a vector map of an environment. The vector map includes lines that outline the shape of objects in the environment (e.g., shelves on the floor of a warehouse). The system identifies one or more line segments representing the boundary or outline of the objects in the environment using range data acquired by the range sensors. The robotic system can repeat this process at different locations as it moves in the environment. Because of errors and inaccuracies, line segments formed at different locations may not clearly align even when these line segments correspond to the same object. To account for this error, the robotic system match line segments identified at a first location with line segments identified at a second location. The matched line segments can be merged into a line that is stored in the vector map.

Physical models for hierarchical clustering and symbolic inference

Growing community of inventors

Burlington, MA, United States of America

Jong Jin Park