Design and Experimental Validation of a LiDAR–SLAM and Model Predictive Control Based Autonomous Robot for Voice-Assisted Hospital Logistics
Shreyan Majumdar, Saumya Ezhil, Sreejon Mondal, Naren Medarametla, Pradeep N
Proceedings of the ASME 2026 International Mechanical Engineering Congress and Exposition (IMECE-INDIA2026)
Healthcare institutions increasingly face workforce shortages and rising logistical demands, placing substantial strain on nursing staff who must balance direct patient care with routine transportation and inventory management tasks. This paper presents the design and experimental validation of a voice-assisted autonomous transportation and storage robot for hospital environments, aimed at improving internal logistics efficiency and operational safety. The proposed system integrates LiDAR-based Visual-SLAM for real-time environment mapping with a Model Predictive Control (MPC)–driven navigation framework to enable robust autonomous mobility in dynamic, human-shared spaces. A two-dimensional occupancy map is generated from LiDAR range measurements and serves as the foundational representation of the environment. Obstacle detection and free-space segmentation are performed online, while future collision risk is estimated through short-horizon motion prediction of surrounding agents. These perception outputs are embedded directly into the MPC formulation, allowing real-time computation of dynamically feasible and collision-free trajectories under kinematic constraints. Continuous replanning is executed at each control cycle, yielding average end-to-end navigation latency on the order of tens of milliseconds and MPC solve times below 50 ms, enabling responsive operation in cluttered environments. A smart storage and inventory framework with voice-based human–robot interaction is additionally developed, enabling natural-language commands for medicine retrieval, deposition, and delivery. Experimental evaluation in representative hospital-like scenarios demonstrates task success rates exceeding 90%, mean path-tracking errors within a few centimeters, and voice-command recognition accuracy above 90%. The proposed platform highlights the effectiveness of tightly integrated perception, predictive control, and human-centric interfaces for scalable and safe robotic logistics in healthcare facilities.
