Real-time predictor of human movement in shared workspaces

An apparatus for predicting arm movements of a human collaborating with a robot in a collaboration workspace, the apparatus comprising

• an observation circuit configured to determine a current kinematic state of the human, determine occupancy information about objects in the collaboration workspace, and determine a goal of the human based on the occupancy information and the current kinematic state;
• a geometric planner circuit configured to determine a possible trajectory for the human based on the goal and the current kinematic state;
• a short-horizon circuit configured to:
• model a next kinematic state of the human at a next time interval based on a matrix of coefficients that define how much each one of a set of previously observed kinematic states contributes to the next kinematic state;
• and
• determine a short-horizon trajectory for the human based on the set of previously observed kinematic states of the human towards the goal;
• a controller circuit configured to:
• determine a likelihood that the human will follow the possible trajectory, wherein the likelihood is based on the short-horizon trajectory, the goal, and the current kinematic state;
• and
• generate a movement instruction to control movement of the robot based on the likelihood.

A non-transitory computer readable medium, comprising instructions which, if executed, cause one or more processors to:

• determine a current kinematic state of a collaborator in a workspace shared with a robot;
• determine a goal of the collaborator based on occupancy information about objects in the workspace;
• determine a possible trajectory for the collaborator based on the goal and the current kinematic state;
• determine a short-horizon trajectory for the collaborator based on previously observed kinematic states of the collaborator towards the goal;
• evaluate a multivariate normal distribution centered on each trajectory point of the possible trajectory;
• determine a likelihood that the collaborator will follow the possible trajectory based on the short-horizon trajectory, the goal, the current kinematic state, the multivariate normal distribution;
• and
• generate a movement instruction to control movement of the robot based on the likelihood that the collaborator will follow the possible trajectory.