Control command based adaptive system and method for estimating motion parameters of differential drive vehicles

A system, comprising:

• a memory;
• one or more communication interfaces;
• and
• one or more hardware processors coupled to said memory through said one or more communication interfaces, wherein said one or more hardware processors are configured to:
• input one or more time synchronized command input signals to a moving differential drive vehicle, wherein the one or more time synchronized command input signals correspond to providing a plurality of control commands for different movements of the moving differential drive vehicle;
• determine, based on the one or more time synchronized command input signals, a corresponding change in one or more motion parameters of the moving differential drive vehicle;
• train a non-linear model using the one or more time synchronized command input signals and the determined corresponding change in the one or more motion parameters, wherein
• the non-linear model is a Hammerstein wiener model,
• the Hammerstein-Wiener model comprises an input non-linearity block, a linear block, and an output non-linearity block, and
• the input non-linearity block of the Hammerstein-Wiener model captures nonlinearity present in input data which is referred as input nonlinearity and represented by a non-linear function, and
• the input data is a unit step time synchronized command input signal represented as u(n)=c1;
• generate, using the trained non-linear model, an experimental model by computing a mapping function between the one or more time synchronized command input signals and the one or more motion parameters of the moving differential drive vehicle;
• estimate, using the generated experimental model, a value of the one or more motion parameters for one or more incoming time synchronized command input signals;
• perform, using a plurality of data obtained from one or more sensors, a validation of the estimated value of the one or more motion parameters, wherein
• the one or more sensors used for validation include position sensors and velocity sensors, and
• the validation is performed by computing an error between the plurality of data obtained from the one or more sensors and measurements of one or more motion parameters obtained from a ground truth system;
• and
• adaptively update the experimental model to estimate an updated value of the one or more motion parameters based on the validation, wherein the step of adaptively updating the experimental model is performed using the computed error and a gain param.

One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• inputting, via the one or more hardware processors, one or more time synchronized command input signals to a moving differential drive vehicle, wherein the one or more time synchronized command input signals correspond to providing a plurality of control commands for different movements of the moving differential drive vehicle;
• determining, based on the one or more time synchronized command input signals, a corresponding change in one or more motion parameters of the moving differential drive vehicle;
• training a non-linear model using the one or more time synchronized command input signals and the determined corresponding change in the one or more motion parameters, wherein
• the non-linear model is a Hammerstein wiener model,
• the Hammerstein-Wiener model comprises an input non-linearity block, a linear block, and an output non-linearity block, and
• the input non-linearity block of the Hammerstein-Wiener model captures nonlinearity present in input data which is referred as input nonlinearity and represented by a non-linear function, and
• the input data is a unit step time synchronized command input signal represented as u(n)=c1;
• generating, using the trained non-linear model, an experimental model by computing a mapping function between the one or more time synchronized command input signals and the one or more motion parameters of the moving differential drive vehicle;
• estimating, using the generated experimental model, a value of the one or more motion parameters for one or more incoming time synchronized command input signals;
• performing, using a plurality of data obtained from one or more sensors, a validation of the estimated value of the one or more motion parameters, wherein
• the one or more sensors used for validation include position sensors and velocity sensors, and
• the validation is performed by computing an error between the plurality of data obtained from the one or more sensors and measurements of one or more motion parameters obtained from a ground truth system;
• and
• adaptively updating the experimental model to estimate an updated value of the one or more motion parameters based on the validation, wherein the step of adaptively updating the experimental model is performed using the computed error and a gain parameter.