Understanding Sensor Fusion and Tracking, Part 4Understanding Sensor Fusion and Tracking, Part 4

The importance of sensor fusion is growing as more applications require the combination of data from different sensor inputs. Self-driving cars, radar tracking systems, and the IoT are some key applications.

In the fourth installment of this video series, Matlab looks at single-object tracking. The video examines how an IMM (interacting multiple model) filter can be used to track an objcct that goes through distinct maneuvers of constant velocity, constant turn, and constant acceleration. The multiple model algorithms allow for more predictive modeling of object behavior than a single model, with less residual error.

The video show that with the IMM  method, the overall normalized distance is lower for all three maneuvers─constant velocity, constant turn, and constant acceleration─than with the single model results. By building a model for each expected motion and setting up an IMM to blend the results, a more accurate profile of expected motion behavior can be generated.

You can view the video here.