Gm Global Technolgoy Operations, Inc.

Mitsubishi Electric Research Laboratories, Inc.

Michigan State University

United States of America, As Represented by the Secretary of Agriculture

Zhaojian Li

Julian R Verdejo

Uros Kalabic

Abraham Goldsmith

Renfu Lu

Michael E Deffenbaugh

Kyle Lammers

A three-dimensional stereo vision camera communicates the location of a targeted fruit (preferably an apple) to a controller. The controller directs a vacuum selection tube to the location of the targeted apple. The selection tube has a gripping orifice that has a smaller diameter than the apple so that the vacuum suction from the selection tube grips the apple. The selected fruit is then rotated and pulled away from the tree canopy by the selection tube until the apple detaches from the canopy. The detached apple is then gently placed in a collection bin for later sorting and processing.

System and method for harvesting fruit

A system for controlling an operation of a machine subject to state constraints in continuous state space of the machine and subject to control input constraints in continuous control input space of the machine is provided. The apparatus includes an input interface to accept data indicative of a state of the machine, a memory configured to store an optimization problem for computing the safety margin of a state and action pair satisfying the state constraints and a control policy mapping the state of the machine within a control invariant set (CIS) to a control input satisfying the control input constraints, and a processor configured to iteratively perform a reinforcement learning (RL) algorithm to jointly control the machine and update the control policy.

Constraint adaptor for reinforcement learning control

An engine control system for a vehicle, includes a model predictive control (MPC) module that identifies sets of possible target values based on an engine torque request, determines predicted operating parameters for the sets of possible target values, determines cost values for the sets of possible target values, selects one of the sets of possible target values based on the cost values; and sets target values based on the possible target values of the selected one of the sets; and a first actuator module that controls a first engine actuator based on a first one of the target values. The MPC module determines at least one of the predicted operating parameters at a future point in time based on a predicted value of engine rpm, which is determined based on a plurality of recent engine rpm measurements.

Use of predicted RPM in multi-variable engine control with least-square techniques

Growing community of inventors

Ann Arbor, MI, United States of America

Zhaojian Li