Batch-reactor input profiles are normally optimized off-line, given the state model and the initial conditions. However, online adaption of the input profile is often necessary in order to preserve optimality in the face of deviation of the batch run from the pre-computed state trajectory Online reoptimization of the future input profile requires prohibitively large real-time computing power, even when techniques such as orthogonal collocation are used to reduce the computational effort. A mechanism for online correction of the pre-optimized input profile, requiring as little online computing power as a P-controller, is examined in this work. For a simulated batch reactor with initial conditions deviating from nominal, the correction procedure yields results close to the theoretical optimum attainable through computationally unattractive online reoptimization.