# US 2020/0293010 A1 · Model-Based Control with Uncertain Motion Model

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2020/0293010 A1
- Document ID: US20200293010A1
- Status: Published application
- Kind: A1
- Application number: 16/298,271
- Filed: 2019-03-11
- Published: 2020-09-17
- Earliest claimed: 2019-03-11
- Assignees: Mitsubishi Electric Research Laboratories, Inc.
- Applicants: Mitsubishi Electric Research Laboratories, Inc.
- Inventors: Karl Berntorp
- Web page: https://patentgrep.com/patent/US20200293010A1

## CPC

- G05B 13/048: {using a predictor}
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G05B 2219/41146: Kalman filter
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A system is controlled using particle filter executed to estimate weights of a set of particles based on fitting of the particles into a measurement model, wherein a particle includes a motion model of the system having an uncertainty modeled as a Gaussian process over possible motion models of the system and a state of the system determined with the uncertainty of the motion model of the particle, wherein a distribution of the Gaussian process of the motion model of one particle is different from a distribution of the Gaussian process of the motion model of another particle. Each execution of the particle filter updates the state of the particle according to a control input to the system and the motion model of the particle with the uncertainty and determines particle weights by fitting the state of the particle in the measurement model subject to measurement noise.

## Claims

### Claim 1

An apparatus for controlling a system, comprising:
a memory configured to store a particle filter configured to estimate weights of a set of particles based on fitting of the particles into a measurement model, wherein a particle includes a motion model of the system having an uncertainty modeled as a Gaussian process over possible motion models of the system and a state of the system determined with the uncertainty of the motion model of the particle, wherein a distribution of the Gaussian process of the motion model of one particle is different from a distribution of the Gaussian process of the motion model of another particle, wherein each execution of the particle filter updates the state of the particle according to a control input to the system and the motion model of the particle with the uncertainty, and determines the weight of the particle by fitting the state of the particle in the measurement model subject to measurement noise; and
a processor configured to
execute the particle filter to produce the weight of each particle and the state of each particle determined with the uncertainty of the motion model of each particle;
update the distribution of the Gaussian process of a particle based on a difference between the state of the particle determined with the uncertainty of the motion model of the particle and a state of the particle determined without the uncertainty of the motion model of the particle;
determine a current estimate of the motion model as a weighted combination of the motion models of the particles weighted according to the weights of the particles;
determine a current estimate of the state of the system as a weighted combination of the states of the particles weighted according to the weights of the particles; and
execute a controller configured to generate a value of the control input based on the current estimate of the motion model and the current estimate of the state of the system and control the system using the value of the control input.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein the Gaussian process of each particle is parameterized as a weighted combination of a set of basis functions, wherein the distribution of the Gaussian process is a Student-t distribution having a mean defining weights of the weighted combination and a scale defining the uncertainty of the motion model of the particle.

### Claim 3 (depends on 2)

The apparatus of claim 2, wherein the weights of the basis functions are modeled as a weight matrix, wherein the weight matrix is Matrix-Normal distributed.

### Claim 4 (depends on 2)

The apparatus of claim 2, wherein the execution of the particle filter samples from the Student-t distribution to produce the motion model of the system with the uncertainty.

### Claim 5 (depends on 2)

The apparatus of claim 2, wherein the execution of the particle filter generates a random number sample according to the Student-t distribution, determines the state of the particle without the uncertainty of the motion model of the particle, and updates the state of the particle without the uncertainty of the motion model of the particle according to the random number to produce the state of the particle with the uncertainty of the motion model of the particle.

### Claim 6 (depends on 2)

The apparatus of claim 2, wherein the state of the particle without the uncertainty of the motion model is determined by updating the state with the motion model defined by the mean of the Student-t distribution.

### Claim 7 (depends on 1)

The apparatus of claim 1, wherein the controller determining the control input is a model based controller.

### Claim 8 (depends on 7)

The apparatus of claim 7, wherein the model based controller adapts its control input according to the uncertainty estimate of the motion model and the uncertainty estimate of the state.

### Claim 9 (depends on 1)

The apparatus of claim 1, wherein the processor for each time step in the motion data executes the particle filter configured to
determine a set of particles, each particle represents the state of the system determined with a sample of the motion model drawn from a probability distribution of the motion models represented as the Gaussian process; and
compare the state of each particle from the set of particles with measurements of the state according to the measurement model to determine a weight of each particle representing an error between the state of the particle and a state trajectory indicated by the measurements.

### Claim 10 (depends on 1)

The apparatus of claim 1, wherein the controlled system is a vehicle, wherein the control inputs include commands specifying values of one or combination of a steering angle of the wheels of the vehicle and a rotational velocity of the wheels, and wherein the measurements include values of one or combination of a rotation rate of the vehicle and an acceleration of the vehicle, each state includes a velocity and a heading rate of the vehicle, such that the motion model relates the value of the control inputs to a first value of the state of the vehicle through dynamics of the vehicle at consecutive time steps, and the measurement model relates the value of the measurement to a second value of the state of the vehicle at the same time step.

### Claim 11

A method for controlling a system, wherein the method uses a processor coupled to a memory storing a particle filter configured to estimate weights of a set of particles based on fitting of the particles into a measurement model, wherein a particle includes a motion model of the system having an uncertainty modeled as a Gaussian process over possible motion models of the system and a state of the system determined with the uncertainty of the motion model of the particle, wherein a distribution of the Gaussian process of the motion model of one particle is different from a distribution of the Gaussian process of the motion model of another particle, wherein each execution of the particle filter updates the state of the particle according to a control input to the system and the motion model of the particle with the uncertainty, and determines the weight of the particle by fitting the state of the particle in the measurement model subject to measurement noise, wherein the processor is coupled with stored instructions implementing the method, wherein the instructions, when executed by the processor carry out steps of the method, comprising:
executing the particle filter to produce the weight of each particle and the state of each particle determined with the uncertainty of the motion model of each particle;
updating the distribution of the Gaussian process of a particle based on a difference between the state of the particle determined with the uncertainty of the motion model of the particle and a state of the particle determined without the uncertainty of the motion model of the particle;
determining a current estimate of the motion model as a weighted combination of the motion models of the particles weighted according to the weights of the particles;
determining a current estimate of the state of the system as a weighted combination of the states of the particles weighted according to the weights of the particles; and
executing a controller configured to generate a value of the control input based on the current estimate of the motion model and the current estimate of the state of the system and control the system using the value of the control input.

### Claim 12 (depends on 11)

The method of claim 11, wherein the Gaussian process of each particle is parameterized as a weighted combination of a set of basis functions, wherein the distribution of the Gaussian process is a Student-t distribution having a mean defining weights of the weighted combination and a scale defining the uncertainty of the motion model of the particle.

### Claim 13 (depends on 12)

The method of claim 12, wherein the weights of the basis functions are modeled as a weight matrix, wherein the weight matrix is Matrix-Normal distributed.

### Claim 14 (depends on 12)

The method of claim 12, wherein the execution of the particle filter samples from the Student-t distribution to produce the motion model of the system with the uncertainty.

### Claim 15

A non-transitory computer readable storage medium embodied thereon a program executable by a processor for performing a method, the medium stores a particle filter configured to estimate weights of a set of particles based on fitting of the particles into a measurement model, wherein a particle includes a motion model of the system having an uncertainty modeled as a Gaussian process over possible motion models of the system and a state of the system determined with the uncertainty of the motion model of the particle, wherein a distribution of the Gaussian process of the motion model of one particle is different from a distribution of the Gaussian process of the motion model of another particle, wherein each execution of the particle filter updates the state of the particle according to a control input to the system and the motion model of the particle with the uncertainty, and determines the weight of the particle by fitting the state of the particle in the measurement model subject to measurement noise, the method comprising:
executing the particle filter to produce the weight of each particle and the state of each particle determined with the uncertainty of the motion model of each particle;
updating the distribution of the Gaussian process of a particle based on a difference between the state of the particle determined with the uncertainty of the motion model of the particle and a state of the particle determined without the uncertainty of the motion model of the particle;
determining a current estimate of the motion model as a weighted combination of the motion models of the particles weighted according to the weights of the particles;
determining a current estimate of the state of the system as a weighted combination of the states of the particles weighted according to the weights of the particles; and
executing a controller configured to generate a value of the control input based on the current estimate of the motion model and the current estimate of the state of the system and control the system using the value of the control input.
