Various embodiments of the technology described herein generally relate to systems and methods for trajectory optimization with machine learning techniques. More specifically, certain embodiments relate to using neural networks to quickly predict optimized robotic arm trajectories in a variety of scenarios. Systems and methods described herein use deep neural networks to quickly predict optimized robotic arm trajectories according to certain constraints. Optimization, in accordance with some embodiments of the present technology, may include optimizing trajectory geometry and dynamics while satisfying a number of constraints, including staying collision-free and minimizing the time it takes to complete the task.

Methods, apparatuses, systems, and computer program products for an improved anti-sway control system and adjustable end effector for a robotic arm are provided. An example method includes determining at least one of a size, shape or orientation of a package to be picked up by an end effector of a robotic arm and facilitating adjusting a position of a suction cup on the end effector, wherein the position is determined based on the at least one determined size, shape, or orientation of the package. The method further includes facilitating grasping the package with the end effector and facilitating movement of the end effector via a robotic joint to reduce force on the suction cup by the package due to an acceleration of the package due to movement of the robotic arm.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for robot plan online adjustment. A method includes receiving an initial plan for performing a particular task with a robot having a sensor. The initial plan defines an initial path having a plurality of waypoints and a footprint space for performing the particular task. The method includes receiving a sensor input from the sensor, and determining from the sensor input that the initial plan requires modification to complete the particular task. The method includes evaluating a plurality of alternative paths, including rejecting one or more alternative paths having any alternative waypoints that are outside the footprint space defined by the initial plan. The method includes selecting an alternative path having alternative waypoints that are all within the footprint space defined by the initial plan. The method includes following, by the robot, the alternative path.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for performing robot plan online adjustment. A method includes receiving an initial plan for performing a particular task with a robot having a sensor. The initial plan defines an initial path having a plurality of waypoints. Each waypoint is associated with a target position and a target velocity. The method includes generating an alternative path from the initial path. Generating an alternative path includes generating a plurality of alternative paths including performing respective modifications to one or more waypoints in the initial plan, evaluating each alternative path according to a simulated total time duration required for the robot to traverse the alternative path, and selecting an alternative path having a total time duration that is less than a total time duration of the initial plan.

A trajectory control device includes: a contact sensor that can contact side surfaces of a workpiece; an actuator that moves a trajectory tracking member and the contact sensor; and a trajectory controller that calculates XY coordinates of a trajectory on the workpiece that is placed in an arbitrary position, by transforming XY coordinates of the trajectory on the workpiece in a reference position, based on positional information about the side surfaces of the workpiece in the reference position and positional information about the side surfaces of the workpiece placed in the arbitrary position. The positional information about the side surfaces of the workpiece placed in the arbitrary position is obtained by the contact sensor.

A system for robot and human collaboration. The system comprises: a multi-axis robot; one or more torque sensors, each torque sensor being configured to measure a torque about a respective axis of the multi-axis robot; and a controller configured to: receive one or more torque measurements taken by the one or more torque sensors; compare the one or more torque measurements or a function of the one or more torque measurements to a threshold value; and control the multi-axis robot based on the comparison.

A rigid-flexible coupling high-accuracy galvo scanning motor comprises: a stator, a rotor rotating relative to the stator, bearing seats and at least two groups of encoders. The rigid-flexible coupling bearings are installed on the rotating shaft of the rotor; each of the rigid-flexible coupling bearings comprises: a rigid bearing and a flexible hinge ring which is elastically deformable, and the flexible hinge ring is fixed in an inner ring of the rigid bearing; the at least two groups of encoders comprise: a first group of encoders and a second group of encoders; the first group of encoders is used to measure a rotation angle of the rotating shaft; and the second group of encoders is used to measure a rotation angle of the inner ring of the rigid bearing. A friction dead zone is avoided through the elastic deformation of the flexible hinge ring, thereby reducing a disturbance bandwidth.

In some aspects, computer-implemented methods for selecting a robotic tool path for a manufacturing processing system to execute a material processing sequence in three-dimensional space can include: providing to a computer-readable product including robotic system data of a robotic tool handling system and workpiece data relating to a processing path of a tool along the workpiece; generating a plurality of possible robotic tool paths to be performed to move the tool along the processing path; identifying one or more obstacles, or an absence of obstacles, associated with the robotic tool paths; comparing robotic tool paths based on a predetermined robotic parameter to be controlled as the tool moves from the start point to the end point; and based on the identified obstacles, determining feasible tool paths, between the start point and the end point that avoid the obstacles, that can be obtained by adjusting the predetermined robotic parameter.

Algorithmic reasoning about a cutting tool assembly's space of feasible configurations can be effectively harnessed to construct a sequence of motions that guarantees a collision-free path for the tool assembly to remove each support structure in the sequence. A greedy algorithm models the motion of the cutting tool assembly through the free-spaces around the intermediate shapes of the part as the free-spaces iteratively reduce in size to the near-net shape to determine feasible points of contact for the cutting tool assembly. Each support beam is evaluated for a contact feature along the boundary of the near-net shape that constitutes a feasible point of contact. If a support beam has at least one feasible configuration at each point, the support beam is deemed accessible and a collection of tool assembly configurations that are guaranteed to be non-colliding but which can access all points of contact of each accessible support beam can be generated.

Faster, less computational intense, and more robust techniques to optimize velocity of robots or portions thereof without violating constraints on acceleration and jerk (derivative of acceleration with respect to time) are described. A nonlinear problem of optimizing velocity without violating acceleration constraints is linearized, and produces acceleration constrained velocity estimates. A nonlinear problem of optimizing velocity without violating jerk constraints in linearized, and produces jerk constrained velocity estimates, and may be feed by the acceleration constrained velocity estimates. Configuration and timing may be generated and provided, e.g., as vectors, to control operation of a robot, robotic appendage or other structure.