A method for planning a manipulation task of an agent, particularly a robot. The method includes: learning a number of manipulation skills wherein a symbolic abstraction of the respective manipulation skill is generated; determining a concatenated sequence of manipulation skills selected from the number of learned manipulation skills based on their symbolic abstraction so that a given goal specification indicating a given complex manipulation task is satisfied; and executing the sequence of manipulation skills.

A movement path drawing device includes an execution time storage unit that stores execution times of respective blocks of control programs, a program analysis unit that creates movement command data by analyzing the control programs, a movement path creation unit that creates the movement paths of the movable parts on the basis of the created movement command data, a drawing execution control unit that performs drawing execution control for drawing movement paths indicating a positional relationship between the movable parts of the plurality of systems of the machine at a predetermined time on the basis of the execution times of the respective blocks and the movement paths of the movable parts, created by the movement path creation unit, and a drawing unit that executes drawing processing for drawing the movable parts of the plurality of systems.

Solutions for multi-robot configurations are co-optimized, to at least some degree, across a set of non-homogenous parameters based on a given set of tasks to be performed by robots in a multi-robot operational environment. Non-homogenous parameters may include two or more of: the respective base position and orientation of the robots, an allocation of tasks to respective robots, respective target sequences and/or trajectories for the robots. Such may be executed pre-runtime. Output may include for each robot: workcell layout, an ordered list or vector of targets, optionally dwell time durations at respective targets, and paths or trajectories between each pair of consecutive targets. Output may provide a complete, executable, solution to the problem, which in the absence of variability in timing, can be used to control the robots without any modification. A genetic algorithm, e.g., Differential Evolution, may optionally be used in generating a population of candidate solutions.

A method for controlling a robot. The method includes providing demonstrations for performing each of a plurality of skills; training from the demonstrations, a robot trajectory model for each skill, each trajectory model is a hidden semi-Markov model having one or more initial states and one or more final states; training, from the demonstrations, a precondition model for each skill comprising, for each initial state, a probability distribution of robot configurations before executing the skill, and a final condition model for each skill comprising, for each final state, a probability distribution of robot configurations after executing the skill; receiving a description of a task, the task includes performing the skills of the plurality of skills in sequence and/or branches; generating a composed robot trajectory model; and controlling the robot according to the composed robot trajectory model to execute the task.

Techniques are disclosed for controlling robotic systems to perform assembly tasks. In some embodiments, a robot control application receives sensor data associated with one or more parts. The robot control application applies a grasp perception model to predict one or more grasp proposals indicating regions of the one or more parts that a robotic system can grasp. The robot control application causes the robotic system to grasp one of the parts based on a corresponding grasp proposal. If the pose of the grasped part needs to be changed in order to assemble the part with one or more other parts, the robot control application determines movements of the robotic system required to re-grasp the part in a different pose. In addition, the robot control application determines movements of the robot system for assembling the part with the one or more other parts based on results of a motion planning technique.

Systems and methods for online augmentation for learned grasping are provided. In one embodiment, a method is provided that includes identifying an action from a discrete action space. The method includes identifying a second set of grasps of the agent utilizing a transition model based on the action and at least one contact parameter. The at least one contact parameter defines allowed states of contact for the agent. The method includes applying a reward function to evaluate each grasp of the second set of grasps based on a set of contact forces within a friction cone that minimizes a difference between an actual net wrench on the object and a predetermined net wrench. The reward function is optimized online using a lookahead tree. The method includes selecting a next grasp from the second set. The method includes causing the agent to execute the next grasp.

Solutions for multi-robot configurations are co-optimized, to at least some degree, across a set of non-homogenous parameters based on a given set of tasks to be performed by robots in a multi-robot operational environment. Non-homogenous parameters may include two or more of: the respective base position and orientation of the robots, an allocation of tasks to respective robots, respective target sequences and/or trajectories for the robots. Such may be executed pre-runtime. Output may include for each robot: workcell layout, an ordered list or vector of targets, optionally dwell time durations at respective targets, and paths or trajectories between each pair of consecutive targets. Output may provide a complete, executable, solution to the problem, which in the absence of variability in timing, can be used to control the robots without any modification. A genetic algorithm, e.g., Differential Evolution, may optionally be used in generating a population of candidate solutions.

A method for planning a manipulation task of an agent, particularly a robot. The method includes: learning a number of manipulation skills wherein a symbolic abstraction of the respective manipulation skill is generated; determining a concatenated sequence of manipulation skills selected from the number of learned manipulation skills based on their symbolic abstraction so that a given goal specification indicating a given complex manipulation task is satisfied; and executing the sequence of manipulation skills.

A robot operable within a 3-D volume includes a gripper movable between an open position and a closed position to grasp any one of a plurality of objects, an articulatable portion coupled to the gripper and operable to move the gripper to a desired position within the 3-D volume, and an object detection system operable to capture information indicative of the shape of a first object of the plurality of objects positioned to be grasped by the gripper. A computer is coupled to the object detection system. The computer is operable to identify a plurality of possible grasp locations on the first object and to generate a numerical parameter indicative of the desirability of each grasp location, wherein the numerical parameter is at least partially defined by the next task to be performed by the robot.

Techniques are disclosed for a trajectory planning of robots, such as collaborative robots (cobots). A controller of a robot may include a path planner, a trajectory generator, and a trajectory controller. The path planner may determine a plurality of waypoints defining a path between an initial pose of the robot and a goal pose of the robot. The trajectory generator may determine a trajectory between the initial pose and the goal pose based on the waypoints and one or more trajectory criterion. The trajectory controller may generate a control signal to control the robot based on the determined trajectory.