Powar, Gayatri. Calculation of collision probability for autonomous vehicles using trajectory prediction. Retrieved from https://doi.org/doi:10.7282/T3F19215
DescriptionThe aim of this thesis is to create a decision making algorithm. The goal would be to check the feasibility of the current maneuver by nding the probability of collision between the subject and target vehicles. These outputs can be used by other Advanced Driver Assistance System (ADAS) features including path planner, lateral control, longitudinal control, etc. We make use of some sensors like camera/radar (simulated data) and fuse these together for better estimation of measurements. With earlier experience with cameras, they are really poor at giving longitudinal distances whereas radars give more accurate measurements longitudinally. Using these measurements about targets ahead in the environment, we predict the trajectories of the obstacles/targets as well as the subject vehicle (autonomous vehicle). The algorithm predicts if it is safe to continue with the current maneuver in the near future for several seconds ahead of time, and makes the decision if the maneuver is possible. The results are obtained using probabilistic approach whether the future trajectories are going to collide. The thesis primarily focuses on target tracking, e cient sensor data fusion and future collision estimation. With the lessons learnt using existing literature an e cient approach is employed. The simulation is performed in PreScan simulator and MATLAB. Enhancements in the sensor data fusion using standby measurements and quasi-decentralized approach to combine measurements to yield improved results and achieve better scalability are proposed and implemented.