State estimation apparatus, state estimation method, integrated circuit, and non-transitory computer-readable storage medium

A state estimation apparatus for estimating an internal state of an observation target, the state estimation apparatus comprising:

• circuitry including at least a processor and a memory, the circuitry configured to:
• obtain first observation data from an observable event at predetermined time intervals;
• obtain second observation data from the observable event at predetermined time intervals;
• store an initial value for generating prior probability distribution data in an initial state;
• in states other than the initial state, perform prediction using, as prior probability distribution data at current time t, posterior probability distribution data obtained at preceding time t−1 indicating a probability distribution of the internal state of the observation target to obtain predictive probability distribution data indicating the probability distribution of the internal state of the observation target at the current time t, wherein the prior probability distribution data is generated based on the initial value during the initial state;
• obtain a first likelihood based on the first observation data;
• obtain a second likelihood based on the second observation data;
• obtain a combined likelihood based on the first likelihood, the second likelihood, first reliability data indicating reliability of the first observation data, and second reliability data indicating reliability of the second observation data;
• obtain posterior probability distribution data indicating the probability distribution of the internal state of the observation target at the current time t from the combined likelihood and the predictive probability distribution data
• obtain the first reliability data;
• obtain the second reliability data, wherein the circuitry obtains:
• first detection accuracy data indicating detection accuracy of the first observation data,
• second detection accuracy data indicating detection accuracy of the second observation data,
• the first reliability data is based on the first detection accuracy data, and
• the second reliability data is based on the second detection accuracy data;
• and
• obtain one or more sensor detection values based on a predetermined physical quantity that is detected by one or more sensors, wherein the circuitry obtains:
• the first reliability data based on the sensor detection value obtained and the first detection accuracy data, and
• the second reliability data based on the sensor detection value obtained and the second detection accuracy data.

A state estimation method for estimating an internal state of an observation target, the method comprising:

• obtaining first observation data from an observable event at predetermined time intervals;
• obtaining second observation data from the observable event at predetermined time intervals;
• storing an initial value for generating prior probability distribution data in an initial state;
• in states other than the initial states, performing prediction using, as prior probability distribution data at current time t, posterior probability distribution data obtained at preceding time t−1 indicating a probability distribution of an internal state of an observation target to obtain predictive probability distribution data indicating a probability distribution of the internal state of the observation target at the current time t, wherein the prior probability distribution data is generated based on the initial value during the initial state;
• obtaining a first likelihood based on the first observation data;
• obtaining a second likelihood based on the second observation data;
• obtaining a combined likelihood based on the first likelihood, the second likelihood, first reliability data indicating reliability of the first observation data, and second reliability data indicating reliability of the second observation data;
• obtaining posterior probability distribution data indicating a probability distribution of the internal state of the observation target at the current time t from the combined likelihood and the predictive probability distribution data;
• obtaining the first reliability data;
• obtaining the second reliability data by obtaining:
• first detection accuracy data indicating detection accuracy of the first observation data,
• second detection accuracy data indicating detection accuracy of the second observation data,
• the first reliability data being based on the first detection accuracy data, and
• the second reliability data being based on the second detection accuracy data;
• and
• obtaining one or more sensor detection values based on a predetermined physical quantity that is detected by one or more sensors by obtaining:
• first corrected reliability data by correcting the first reliability data based on the sensor detection value obtained,
• second corrected reliability data by correcting the second reliability data based on the sensor detection value obtained, and
• the combined likelihood based on the first corrected reliability data and the second corrected reliability data.

A state estimation method for estimating an internal state of an observation target, the method comprising:

• obtaining first observation data from an observable event at predetermined time intervals;
• obtaining second observation data from the observable event at predetermined time intervals;
• storing an initial value for generating prior probability distribution data in an initial state;
• in states other than the initial states, performing prediction using, as prior probability distribution data at current time t, posterior probability distribution data obtained at preceding time t−1 indicating a probability distribution of the internal state of the observation target to obtain predictive probability distribution data indicating the probability distribution of the internal state of the observation target at the current time t, wherein the prior probability distribution data is generated based on the initial value during the initial state;
• obtaining a first likelihood based on the first observation data;
• obtaining a second likelihood based on the second observation data;
• obtaining a combined likelihood based on the first likelihood, the second likelihood, first reliability data indicating reliability of the first observation data, and second reliability data indicating reliability of the second observation data;
• obtaining posterior probability distribution data indicating the probability distribution of the internal state of the observation target at the current time t from the combined likelihood and the predictive probability distribution data;
• obtaining the first reliability data;
• obtaining the second reliability data by obtaining:
• first detection accuracy data indicating detection accuracy of the first observation data,
• second detection accuracy data indicating detection accuracy of the second observation data,
• the first reliability data being based on the first detection accuracy data, and
• the second reliability data being based on the second detection accuracy data;
• obtaining one or more sensor detection values based on a predetermined physical quantity that is detected by one or more sensors by obtaining:
• the first reliability data based on the sensor detection value obtained and the first detection accuracy data, and
• the second reliability data based on the sensor detection value obtained and the second detection accuracy data.