SENSOR INFORMATION PROCESSING AND ADAPTIVE CONTROL FOR REAL-TIME MOTION AUTOMATION OF AUTONOMOUS ROBOTS: MULTI-RATE STATE ESTIMATION WITH DELAYED MEASUREMENTS

Autonomous mobile robots must turn fast, drifting proprioceptive data and slow, delayed, outlier-prone pose fixes into a state estimate that a controller can use while payload and friction change. This simulation study asks what delayed-measurement replay, innovation gating and a projection-based adaptive term add to a baseline PID loop for a differential-drive robot. A multi-rate extended Kalman filter (EKF) fuses 100 Hz encoders and gyroscope with 10 Hz pose fixes (60 ms latency, 2 % outliers); boundedness of the adaptive term is analysed for a scalar channel. A figure-eight path with a payload step was simulated with 40 paired seeds per condition, with ablations, an obstacle scenario and timing of the Python code. In simulation, the full pipeline gave a position RMSE of 0.024 m, against 0.372 m for encoders and gyroscope only; omitting delay handling or gating raised it by 45 % and 57 %. All 499 outlier fixes were rejected, with 1.3 % of valid fixes, and the filter was overconfident (mean NEES 10.2, expected 3). The adaptive term shortened speed recovery after the step from 1.02 s to 0.68 s but left the tracking error unchanged (0.038 m) and raised actuator saturation from 2.4 % to 6.3 %. The safety filter prevented contact under the full pipeline (0 of 40 runs) but not with encoder-gyroscope estimates (11 of 40). Compute time averaged 0.4 ms per cycle. The plant model is simple, the boundedness argument does not cover the implemented system, and hardware validation is pending.