Business Context
AutoTrack, a company specializing in autonomous vehicle technology, aims to enhance the tracking accuracy of its vehicles in real-time. The ability to predict and smooth the position of vehicles based on noisy sensor data is crucial for safe navigation and operational efficiency.
Dataset
| Feature Group | Count | Examples |
|---|
| Sensor Data | 100K | x_position, y_position, speed, acceleration, timestamp |
| Noise Metrics | 5 | sensor_noise_x, sensor_noise_y, sensor_noise_speed, sensor_noise_acceleration |
- Size: 100K observations with 5 features
- Target: Continuous — estimated position (x, y) of the vehicle
- Class balance: Not applicable as this is a regression problem
- Missing data: 5% missing in sensor readings due to communication errors
Requirements
- Implement a Kalman filter to estimate the vehicle's position based on the noisy sensor data.
- Provide a method to visualize the estimated trajectory versus the actual trajectory.
- Evaluate the performance of the Kalman filter using RMSE and compare it against a simple moving average approach.
- Discuss the advantages of using the Kalman filter in this context.
Constraints
- The solution must run in real-time with a latency of less than 100ms.
- The model should be able to handle missing data gracefully without significant performance degradation.
