17896778. INTERPRETABLE KALMAN FILTER COMPRISING NEURAL NETWORK COMPONENT(S) FOR AUTONOMOUS VEHICLES simplified abstract (Zoox, Inc.)
Contents
- 1 INTERPRETABLE KALMAN FILTER COMPRISING NEURAL NETWORK COMPONENT(S) FOR AUTONOMOUS VEHICLES
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 INTERPRETABLE KALMAN FILTER COMPRISING NEURAL NETWORK COMPONENT(S) FOR AUTONOMOUS VEHICLES - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Potential Commercial Applications
- 1.9 Possible Prior Art
- 1.10 Unanswered Questions
- 1.11 Original Abstract Submitted
INTERPRETABLE KALMAN FILTER COMPRISING NEURAL NETWORK COMPONENT(S) FOR AUTONOMOUS VEHICLES
Organization Name
Inventor(s)
John Bryan Carter of Upton MA (US)
Francesco Papi of Oakland CA (US)
Qian Song of San Mateo CA (US)
Zachary Sun of San Francisco CA (US)
INTERPRETABLE KALMAN FILTER COMPRISING NEURAL NETWORK COMPONENT(S) FOR AUTONOMOUS VEHICLES - A simplified explanation of the abstract
This abstract first appeared for US patent application 17896778 titled 'INTERPRETABLE KALMAN FILTER COMPRISING NEURAL NETWORK COMPONENT(S) FOR AUTONOMOUS VEHICLES
Simplified Explanation
The abstract describes a modified Kalman filter that incorporates neural networks to enhance or replace certain components while maintaining human interpretability. The neural networks can address bias in measurement data, unknown controls in predicting object states, different sensor data training for an ensemble, determining the Kalman gain, and utilizing multiple Kalman filters with neural networks for independent estimated states.
- Neural networks integrated into a modified Kalman filter
- Neural networks address bias in measurement data, unknown controls, different sensor data training, Kalman gain determination, and multiple Kalman filters for independent estimated states
- Maintain human interpretability of the filter's inner functions
Potential Applications
The technology can be applied in various fields such as autonomous vehicles, robotics, aerospace, and industrial automation for accurate state estimation and control.
Problems Solved
1. Improved accuracy in state estimation by addressing bias and unknown controls. 2. Enhanced adaptability to different sensor data through neural network ensembles.
Benefits
1. Increased reliability and robustness in state estimation. 2. Enhanced performance in dynamic and uncertain environments. 3. Improved interpretability of the filter's inner workings.
Potential Commercial Applications
Optimal for use in autonomous vehicles for precise localization and navigation, in robotics for accurate motion planning, in aerospace for flight control systems, and in industrial automation for efficient process control.
Possible Prior Art
There may be prior art related to the integration of neural networks with Kalman filters for state estimation and control in various applications. Research papers or patents in the fields of robotics, autonomous systems, and control theory may have similar approaches.
Unanswered Questions
How does the neural network ensemble training process work in practice?
The article does not delve into the specific methodologies or algorithms used to train the neural network ensemble based on different sensor data.
What are the computational requirements for implementing multiple neural networks in the modified Kalman filter?
The article does not provide information on the computational resources needed to incorporate and run multiple neural networks within the filter.
Original Abstract Submitted
A modified Kalman filter may include one or more neural networks to augment or replace components of the Kalman filter in such a way that the human interpretability of the filter's inner functions is preserved. The neural networks may include a neural network to account for bias in measurement data, a neural network to account for unknown controls in predicting a state of an object, a neural network ensemble that is trained differently based on different sensor data, a neural network for determining the Kalman gain, and/or a set of Kalman filters including various neural networks that determine independent estimated states, which may be fused using Bayesian fusion to determine a final estimated state.
