The Kalman filter
In March 1960 R. E. Kalman published A New Approach to Linear Filtering and Prediction Problems in the Journal of Basic Engineering (Transactions of the ASME, volume 82, issue 1, pages 35-45). The classical filtering and prediction problem is re-examined through the state-transition description of a dynamic system: a nonlinear equation is derived for the covariance of the optimal estimation error, and from its solution the coefficients of the optimal linear filter follow without further calculation; the method applies unchanged to stationary and nonstationary statistics.
Why it matters
A recursive estimator appeared that at every step combines a motion model with a noisy measurement and carries a measure of its own uncertainty. Every record in this atlas about localisation and mapping stands on it: Smith and Cheeseman merge estimates with its update equations in 1986, and MonoSLAM in 2003 holds the camera pose and the landmarks in one state vector of an extended Kalman filter.
The record rests on the publisher's abstract, because the full text is paywalled. The abstract names three results: the formulation and methods of solution apply without modification to stationary and nonstationary statistics and to growing-memory and infinite-memory filters; a nonlinear difference or differential equation is derived for the covariance matrix of the optimal estimation error, from whose solution the coefficients of the optimal linear filter are obtained without further calculations; and the filtering problem is shown to be the dual of the noise-free regulator problem. The method is applied to two well-known problems. What the record does not claim: that the paper uses the words Kalman filter (the name came later), that it mentions robots or navigation, or any figure beyond the volume and pages. The paper is not a robotics paper; it sits in this topic as the foundation of the state-estimation line without which the records of 1986, 1991 and 2003 do not exist.