A Simple Yet Effective Obstacle Avoider For The Tara Autonomous Car

We present a simple yet effective obstacle avoider for the Intelligent and Autonomous Robotic Automobile (IARA). At each or several motion planning cycles, the IARA's obstacle avoider firstly receives as input an updated map of the environment around the car, the current car's state relative to the map, and a trajectory from the current car's state to the next goal state. Secondly, the obstacle avoider simulates the trajectory. Finally, if the trajectory crashes into an obstacle, then the obstacle avoider decreases the linear velocity commands of the trajectory to prevent the accident. To evaluate the performance of the obstacle avoider, we executed experiments with the IARA's simulator using a real-world sensor data log, which was acquired in the campus of the Universidade Federal do Espirito Santo (UFES). We also carried out experiments with IARA itself, which was driven autonomously on a parking lot of the UFES. Experimental results showed that the obstacle avoider, together with the motion planner, allows the IARA to go from objective to objective safely. In fact, in all the experiments executed with the IARA for about one year, the obstacle avoider operated successfully.