Sensor based motion planning and control of a mobile robot.
Author
Madinali, Oghuz
Term
4. semester
Education
Publication year
2025
Submitted on
2025-06-03
Pages
44
Abstract
Denne afhandling udvikler og afprøver en sensorbaseret ramme for bevægelsesplanlægning og kontrol af autonome, mobile robotter i dynamiske miljøer med både bevægelige forhindringer og andre robotter. Løsningen bygger på ikke-lineær modelprædiktiv regulering (NMPC) implementeret i CasADi, som forudsiger robotbevægelse og samtidig respekterer kinematiske og sikkerhedsmæssige begrænsninger. Lidar-data indgår løbende i optimeringen for at opdage forhindringer og håndhæve sikkerhedsafstande, og et analytisk afledt feedbackled forbedrer robustheden. Rammeværket evalueres i Webots med tre TIAGo base-robotter med differentialdrev-kinematik, hvor hver robot navigerer mod sit mål og koordinerer med naboer gennem begrænsningsbaserede interaktioner. Simulationerne viser pålidelig målopnåelse, glatte baner og decentral, kollisionsfri drift. Samlet peger resultaterne på, at NMPC kombineret med ombord sensorer udgør en praktisk og robust løsning til multi-robot-navigation i dynamiske, delvist ukendte miljøer. Fremtidigt arbejde omfatter SLAM-integration, hardwareimplementering og udbygget multi-agent-kommunikation.
This thesis develops and evaluates a sensor-based motion planning and control framework for autonomous mobile robots operating in dynamic environments with moving obstacles and other robots. The approach centers on nonlinear model predictive control (NMPC) implemented in CasADi, which predicts robot motion while respecting kinematic and safety constraints. Real-time lidar data are integrated into the optimization to detect obstacles and enforce separation, and an analytically derived feedback term is added to increase robustness. The framework is assessed in Webots with three TIAGo base robots using differential-drive kinematics; each robot navigates toward its goal while coordinating with nearby agents through constraint-based interactions. Simulations demonstrate reliable goal achievement, smooth trajectories, and decentralized, collision-free operation. Together, the results indicate that NMPC combined with onboard sensing offers a practical and robust solution for multi-robot navigation in dynamic, partially unknown environments. Proposed future directions include integrating SLAM, deploying on real hardware, and enabling richer multi-agent communication.
[This abstract was generated with the help of AI]
Documents