Real-time obstacle avoidance performed by an autonomous vehicle throughout a smooth trajectory using an electronic stick

The CONTROLAB AGV is an electric tricycle vehicle that, following issues from client stations connected to the network office, moves autonomously through an office environment with a known floorplan and uses an electronic stick, which consists of infrared sensors, to detect and avoid unknown obstacles. This paper presents the global real-time solution for the CONTROLAB AGV to move smoothly within an adaptive grided potencial field which changes each time the electronic stick detects a collision point. This modification is simply performed by the addition of previously calculated potential field values on a grid which represents the room working area. The trajectory between two calculated field values is a continuous-curvature path given by a fifth order polynomial function. The interpolating conditions of the fifth order polynomial function have to fit in with the AGV dynamics so as to allow complex motions to be performed. The real-time AGV motion system consists of periodic and aperiodic tasks released by external clocked events. The system is programmed using concurrent processes and threads to improve its robustness and efficiency. The video of the AGV in motion can be seen in the web page www.nce.ufrj.br/controlab.