Apparatuses, systems, and techniques for determining whether collisions will occur in potential paths of an object within a scene. In at least one embodiment, one or more neural networks determine whether collisions will occur in potential paths of an object within a scene based at least in part on point cloud data of the object and the scene.

A robotic imaging system includes a camera configured to obtain one or more images of a target site. A robotic arm is operatively connected to the camera, the robotic arm being adapted to selectively move the camera in a movement sequence. A force-based sensor is configured to detect and transmit sensor data related to at least one of force and/or torque imparted by a user for moving the camera. The system includes a controller configured to receive the sensor data. The controller has a processor and tangible, non-transitory memory on which instructions are recorded. The controller is adapted to selectively execute a collision avoidance mode, including applying a respective correction force to modify the movement sequence when the camera and/or the robotic arm enter a predefined buffer zone.

A trajectory generating method includes a first generating process of generating a plurality of trajectories between a start teaching point and a target teaching point, an evaluation process of evaluating a motion of the robot arm on each trajectory to calculate an evaluation value of each trajectory, a selection process of selecting one of the plurality of trajectories based on calculated evaluation values, and an update process of updating the trajectory by repeating the processes of generating a plurality of new trajectories by changing a selected trajectory in the selection process, of calculating an evaluation value of a motion of the robot arm on each changed trajectory and of selecting a trajectory based on calculated evaluation values.

Disclosed techniques for graph generation for online path planning offer multiple advantages, such as providing for high-quality motion during online operation of the robot, while reducing the computational burden of graph generation. Achieving these competing goals involves reducing the dimensionality of the graph generation problem by performing unconstrained sampling that defines partial robot poses that set values for fewer than all configuration parameters of the robot. The remaining configuration parameters for each sample are then determined in dependence on a distance function that relates the partial pose to one or more reference robot poses that are associated with one or more tasks to be performed by the robot and are provided as inputs to the graph generation. Reference robot poses may be determined automatically based on computer analysis of the robot application or may be user-input values.

A path generator includes an object setter that sets an object model; a device setter that sets a device model as a model of a robot arm; a path generator that generates a path of the robot arm stepwise; and an interference determiner that performs interference determination on the object model and the device model after having moved along the path, based on a distance between the object model and the device model. If it is determined that there is a possibility of interference between the object model and the device model, at least one of the object setter or the device setter increases a density of point groups of the point group model, and performs interference determination by using the point group model with the increased density of the point groups. If determined that that there is no possibility of interference, the path generator generates a next path.