Mobility based on machine-learned movement determination
A mobility augmentation system monitors a user's motor intent data and augments the user's mobility based on the monitored motor intent data. A machine-learned model is trained to identify an intended movement based on the monitored motor intent data. The machine-learned model may be trained based on generalized or specific motor intent data (e.g., user-specific motor intent data). A machine-learned model initially trained on generalized motor intent data may be re-trained on user-specific motor intent data such that the machine-learned model is optimized to the movements of the user. The system uses the machine-learned model to identify a difference between the user's monitored movement and target movement signals. Based on the identified difference, the system determines actuation signals to augment the user's movement. The actuation signals determined can be an adjustment to a currently applied actuation such that the system optimizes the actuation strategy during application.
1 . A method comprising:
applying a plurality of actuation signals produced using a machine-learned movement model and based on an intended movement of a user to one or more augmentation devices worn by the user, the application of the plurality of actuation signals causing an actuated movement of the user, the actuated movement and the intended movement corresponding to a same movement;
computing a difference between feedback signals measuring the actuated movement and a plurality of target movement signals representative of the intended movement;
updating the plurality of actuation signals based on the computed signal difference; and
applying the updated actuation signals to the one or more augmentation devices worn by the user, the application of the updated actuation signals configured to minimize the signal difference and generate a personalized actuation strategy for the intended movement and the user.
2 . The method of claim 1 , further comprising determining the intended movement of the user based on motor intent data received from the user, wherein the motor intent data comprises one or more of electromyography (EMG) data or inertial measurement unit (IMU) data.
3 . The method of claim 1 , further comprising:
monitoring a plurality of movement result signals representative of augmented movement of the user, the application of the plurality of actuation signals to the one or more augmentation devices contributing to the augmented movement;
identifying, using the machine-learned model, a subsequent difference between the plurality of movement result signals and the plurality of target movement signals representative of intended movement of the user;
modifying the plurality of actuation signals based on the subsequent difference; and
applying the plurality of actuation signals to the one or more augmentation devices worn by the user.
4 . The method of claim 1 , further comprising:
collecting neurotypical motor intent data of one or more users from a database, the neurotypical motor intent data corresponding to an activity performed by the one or more users; and
creating a movement template representative of a sequence of muscle activity events corresponding to the activity based on the collected neurotypical motor intent data.
5 . The method of claim 4 , wherein computing a difference between feedback signals measuring the actuated movement and the plurality of target movement signals representative of the intended movement comprises:
determining a movement prediction indicating that the plurality of feedback signals corresponds to the activity using the machine-learned model;
selecting the movement template based on the movement prediction, wherein the sequence of muscle activity events of the movement template is associated with the plurality of target movement signals; and
computing the difference between the feedback signals and the plurality of target movement signals using the selected movement template.
6 . The method of claim 5 , further comprising determining a plurality of actuation signals based on the computed difference by identifying a trigger associated with an actuation strategy based on one or more muscle activity events in the sequence of muscle activity events of the movement template, wherein the actuation strategy comprises an actuation signal of the plurality of actuation signals, an augmentation device of the one or more augmentation devices to which the actuation signal is applied, and a time to apply the actuation signal.
7 . The method of claim 5 , wherein the plurality of target movement signals comprises one or more of kinematic signals, foot plantar pressure signals, or kinetic signals.
8 . The method of claim 1 , further comprising:
collecting neurotypical motor intent data from a coaching user, the neurotypical motor intent data corresponding to an activity performed by the coaching user and monitored using a plurality of sensors worn by the coaching user at one or more locations on the coaching user; and
determining the plurality of target movement signals representative of the intended movement based on the collected neurotypical motor intent data, the plurality of target movement signals characterizing the activity.
9 . The method of claim 1 , wherein the plurality of actuation signals are applied to a muscle group of the user.
10 . The method of claim 1 , further comprising:
collecting neurotypical motor intent data from the user using a first augmentation device of the one or more augmentation devices worn by the user at a first location on the user, the neurotypical motor intent data corresponding to an activity performed by the user; and
determining a plurality of target movement signals representative of the intended movement based on the collected neurotypical motor intent data, the plurality of target movement signals characterizing the activity;
wherein applying the plurality of actuation signals comprises applying the plurality of actuation signals to a second augmentation device of the one or more augmentation devices at a second location on the user, the first and second locations mirrored across a sagittal plane of the body of the user.
11 . The method of claim 1 , further comprising:
determining intent labels based on one or more of foot plantar pressure signals, kinematic signals, and kinetic signals of one or more users from a database;
collecting motor intent data of the one or more users from a database;
preprocessing the collected motor intent data; and
training the machine-learned model using the determined intent labels and the preprocessed motor intent data.
12 . The method of claim 11 , wherein preprocessing the collected motor intent data comprises, for each user of the one or more users:
aligning the motor intent data of the respective user in a temporal domain;
processing the motor intent data using one or more signal processing techniques selected from filtering, averaging, peak-finding, and down-sampling; and
determining kinematic data associated with the processed motor intent data using a biomechanical model.
13 . The method of claim 1 , wherein the plurality of actuation signals comprises one or more of electrical actuation signals, haptic actuation signals, visible signals output from a display of a user device, audio signals, or mechanical actuation signals.
14 . The method of claim 13 , wherein the electrical actuation signals are characterized by a first frequency, a first pulse duration, and a first amplitude for a first period of time and a second frequency, a second pulse duration, and a second amplitude for a second period of time following the first period of time.
15 . The method of claim 1 , wherein applying the plurality of actuation signals comprises actuating an exoskeleton using the plurality of actuation signals.
16 . The method of claim 1 , wherein the one or more augmentation devices comprise an exoskeleton, a plurality of modular electrode straps, or a foot pressure bed.
17 . The method of claim 1 , wherein applying the plurality of actuation signals to the one or more augmentation devices comprises determining a plurality of times to apply the respective plurality of actuation signals based on the intended movement.
18 . The method of claim 17 , further comprising:
monitoring a plurality of movement result signals representative of augmented movement of the user, the application of the plurality of actuation signals to the one or more augmentation devices contributing to the augmented movement;
identifying, using the machine-learned model, a subsequent difference between the plurality of movement result signals and the plurality of target movement signals representative of the intended movement of the user;
modifying the plurality of times based on the subsequent difference; and
applying the actuation signals to the one or more augmentation devices worn by the user at the modified plurality of times.
19 . A mobility improvement system comprising a non-transitory computer-readable storage medium storing instructions for execution and a hardware processor configured to execute the instructions, the instructions, when executed, cause the hardware processor to perform steps comprising:
applying a plurality of actuation signals produced using a machine-learned movement model and based on an intended movement of a user to one or more augmentation devices worn by the user, the application of the plurality of actuation signals causing an actuated movement of the user, the actuated movement and the intended movement corresponding to a same movement;
computing a difference between feedback signals measuring the actuated movement and a plurality of target movement signal representative of the intended movement;
updating the plurality of actuation signals based on the computed signal difference; and
applying the updated actuation signals to the one or more augmentation devices worn by the user, the application of the updated actuation signals configured to minimize the signal difference and generate a personalized actuation strategy for the intended movement and the user.
20 . A method comprising:
determining a plurality of target movement signals based on a movement goal;
monitoring a plurality of movement signals representative of movement of a user, the movement signals determined using sensor data from sensors located at first and second locations on a user;
identifying, based on the movement signals, kinematic patterns associated with the first location on the user;
determining a plurality of actuation signals based on the identified kinematic patterns; and
applying the plurality of actuation signals to a second augmentation device worn by the user at the second location on the user, wherein the plurality of actuation signals are not applied to a first augmentation device worn by the user at the first location on the user.