Videos Prior to 2012 Collaboration with Ford Video of 2012 Experimental Results Driver model with 3d Body pose - Estimation with wearable suite Prediction based on historical and real-time data. Vehicle is controlled based on driver forecasted actions. Driver model with 3d Body pose - Estimation with motion capture system. Collaboration with Ford Model predictive control is used to follow the path of an expert driver. The maneuver consists of bringing the vehicle into a drift and completing a 180 degree turn on ice. Collaboration with Ford  Video of 2013 Experimental Results Nominal MPC with Predictive Driver Model. Goal is to interveen only id lane departure is predicted Collaboration with VOLVO Video of 2014 Experimental Results Collaboration with Hyundai This experiment uses Model Predictive Control (MPC) to keep the the vehicle in the lane while following a reference speed. The figure on the left shows the open-loop paths planned by the controller in blue. The red line is the upcoming road centerline. The figure on the right shows the commanded and measured steering angle. The S-curve is a chicane on the winding road requiring a sharp right turn followed by a sharp left. Midway through the S-curve, the controller plans an aggressive left turn. However, the onboard lower level steering controller is unable to keep up with the MPC command and a lane departure occurs. This is seen by the large discrepancy between the commanded and measured steering values in the figure on the right. This issue has hence been fixed by imposing tighter rate constraints in the MPC formulation and tuning the lower level PID controller. Successful Test (same as above) after proper design was employed. |