Method and system for autonomous vehicle control
AssigneeThe Regents of the University of Michigan
InventorsMatthew Johnson-Roberson, Ram Vasudevan
Life of this patent
Estimating the term…
Abstract
The method for autonomous vehicle control preferably includes sampling measurements, determining refined sensor poses based on the measurements, determining an updated vehicle pose based on the measurements and operation matrix, optionally determining evaluation sensor poses based on the refined sensor poses, optionally updating the operation matrix(es) based on the evaluation sensor poses, and/or any other suitable elements. The system for autonomous vehicle control can include one or more vehicles, one or more sensors, one or more processing systems, and/or any other suitable components.
Claims20
A system for a vehicle comprising:
- a vehicle sensor suite comprising a plurality of sensors;
- and
- a processor configured to, during operation of the vehicle, iteratively:
- calculate sensor poses for each of the plurality of sensors of the vehicle sensor suite based on an operation calibration for the vehicle sensor suite and sensor measurements from the plurality of sensors;
- calculate uncertainties of the sensor measurements for each of the plurality of sensors, the uncertainties represent consistency between sensor measurements for a given sensor;
- compare the uncertainties of the sensor measurements for each of the plurality of sensors at a given timestamp to uncertainties of the sensor measurements at a prior timestamp;
- calculate an evaluation calibration based on the sensor poses and the comparison of uncertainties, wherein the evaluation calibration represents an updated version of the operation calibration at the given timestamp;
- and
- update the operation calibration based on the evaluation calibration being within a threshold difference of a previous evaluation calibration,
- wherein the processor is further configured to flag the vehicle when the evaluation calibration and the previous evaluation calibration satisfy a convergence-failure condition at the timestamp,
- wherein the processor is further configured to calculate a vehicle pose based on the updated operation calibration and the motion of the vehicle and control the vehicle based on the vehicle pose.
A method for a vehicle, comprising:
- during operation of the vehicle, iteratively:
- generating sensor measurements with a plurality of vehicle sensors;
- determining sensor poses for each of the plurality of vehicle sensors based on the sensor measurements and an operation calibration;
- calculating uncertainties for the sensor measurements representing sensor measurement consistency between sensor measurements for a given sensor;
- comparing the uncertainties of the sensor measurements for each of the plurality of sensors at a given timestamp to uncertainties of the sensor measurements at a prior timestamp;
- calculating an evaluation calibration based on the sensor poses and the uncertainties comparison, the evaluation calibration representing an updated version of the operation calibration at the given timestamp;
- selectively updating the operation calibration when the evaluation calibration is within a threshold difference of a previous evaluation calibration;
- calculating a vehicle pose based on the sensor poses and the updated operation calibration;
- controlling the vehicle based on the vehicle pose;
- and
- flagging the vehicle when the evaluation calibration and the previous evaluation calibration satisfy a convergence-failure condition.
Dependencies are inferred from the claim text.