Crew condition estimation system

Abstract

The machine learning model will be designed to meet a wide range of performance requirements when estimating occupant conditions. [Solution] The occupant support system S for estimating the occupant state C of a vehicle V as a moving object comprises a detection device that detects information related to the occupant and the moving object, and an estimation unit 111 that takes the detection signal of the detection device as input and estimates multiple types of occupant states C based on a trained state estimation model M0 which outputs the occupant state C. The state estimation model M0 is composed of a weighted combination of a first model Ma trained based on the personal information of the occupant U himself and a second model Mb trained based on information collected from multiple other occupants U' other than occupant U. The system further comprises a model adjustment unit 115 that changes the weighting in the first model Ma and the second model Mb according to the type of occupant state C to be estimated.

Claims

1. A crew state estimation system that estimates the crew state, which indicates the state of the occupants of a moving vehicle, A detection device mounted on the mobile body for detecting information related to the passenger and the mobile body, The system includes an estimation unit that takes the detection signal from the detection device as input and estimates multiple types of occupant states based on a trained state estimation model that outputs the occupant state, The state estimation model is composed of a weighted combination of a first model learned based on the passenger's own personal information and a second model learned based on information collected from multiple other passengers besides the passenger. The system further includes a model adjustment unit that changes the weighting in the first model and the second model according to the type of occupant condition to be estimated. Crew condition estimation system.

2. In the crew state estimation system described in claim 1, The aforementioned model adjustment unit is Based on the detection signal from the detection device, the estimated status of the occupant's condition is determined. By referring to the estimated conditions in addition to the type of occupant condition, the weighting in the first model and the second model is changed. Crew condition estimation system.

3. In the crew state estimation system described in claim 1, The aforementioned occupant condition includes the physical condition of the passenger and the emotional state of the passenger. The aforementioned model adjustment unit is In estimating the emotions of the passengers, the weights for the second model are set to be smaller than the weights for the first model. In estimating the physical condition of the passenger, a greater weight is given to the second model compared to the estimation of the passenger's emotions. Crew condition estimation system.

4. In the crew state estimation system described in claim 2, The aforementioned estimation status includes information indicating whether or not the moving object is in motion. The aforementioned model adjustment unit is If the moving object is in motion, the weight for the second model is set to zero and the calculation by the second model is omitted. If the moving object is stationary, the weights for the second model are set to non-zero, and the calculations using the second model are performed. Crew condition estimation system.

5. In the crew state estimation system described in claim 2, The aforementioned estimated circumstances include information indicating the passenger's boarding history, The model adjustment unit sets a smaller weight for the second model when the passenger has a substantial flight history compared to when the passenger does not have a substantial flight history. Crew condition estimation system.

6. In the crew state estimation system described in claim 1, The unit further comprises a model update unit for updating the state estimation model, The aforementioned model update unit is The second model is updated by an external computer located outside the mobile unit. The update content from the external computer is reflected at an update timing based on the movement status of the mobile object. Crew condition estimation system.

7. In the crew state estimation system described in claim 6, The model update unit causes the update of the first model to be performed inside the mobile body. Crew condition estimation system.

8. In the crew state estimation system described in claim 7, It is equipped with a notification unit that notifies the aforementioned passenger of information, The notification unit, when updating the state estimation model by the model update unit, notifies the passenger that their personal information has not been communicated outside the mobile body. Crew condition estimation system.

9. In the crew state estimation system described in claim 7, The model update unit, when updating the first model, uses the output from the second model as training data. Crew condition estimation system.

10. In the crew state estimation system described in claim 1, An interactive terminal mounted on the aforementioned mobile device, The system includes a suggestion unit that proposes to the passenger at least one of the following: a crew action performed by the passenger himself or a mobile action performed by the mobile body. The estimation unit estimates the degree of confidence between the passenger and the mobile body, the proposal unit, or the manufacturing brand of the mobile body, based on the passenger's response to the proposal from the proposal unit, as the passenger state. The model adjustment unit, when estimating the occupant state other than the reliability, changes the weighting in the first model and the second model according to the reliability. Crew condition estimation system.

11. In the crew state estimation system described in claim 10, The estimation unit sets the weight for the second model to be smaller as the confidence level increases. Crew condition estimation system.

Images